Matthias
Keil
a,
Alexandre Emmanuel
Wetzel
a,
Kaiyu
Wu
bc,
Elena
Khomtchenko
a,
Jitka
Urbankova
a,
Anja
Boisen
b,
Tomas
Rindzevicius
b,
Ada-Ioana
Bunea
a and
Rafael J.
Taboryski
*a
aDTU Nanolab, National Centre for Nano Fabrication and Characterization, Technical University of Denmark, Ørsted Plads, Building 347, DK-2800 Kgs. Lyngby, Denmark. E-mail: rata@dtu.dk; Tel: +45 45258101
bDTU Health Tech, Department of Health Technology, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark
cNational Key Laboratory of Science and Technology on Micro/Nano Fabrication, Department of Micro/Nano Electronics, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, 200240, P. R. China
First published on 18th February 2021
In this paper, we demonstrate plasmonic color metasurfaces as large as ∼60 cm2 fabricated by deep UV projection lithography employing an innovative combination of resolution enhancement techniques. Briefly, in addition to the established off-axis dipole illumination, double- and cross-exposure resolution enhancement of lithography, we introduce a novel element, the inclusion of transparent assist features to the mask layout. With this approach, we demonstrate the fabrication of relief arrays having critical dimensions such as 159 nm nanopillars or 210 nm nanoholes with 300 nm pitches, which is near the theoretical resolution limit expressed by the Rayleigh criterion for the 248 nm lithography tool used in this work. The type of surface structure, i.e. nanopillar or nanohole, and their diameters can be tailored simply by changing the width of the assist features included in the mask layout. By coating the obtained nanopatterns with thin layers of either Au or Al, we observe color spectra originating from the phenomenon known as localized surface plasmon resonance (LSPR). We demonstrate the generation of color palettes representing a broad spectral range of colors, and we employ finite element modelling to corroborate the measured LSPR fingerprint spectra. Most importantly, the ∼60 cm2 nanostructure arrays can be written in only a few minutes, which is a tremendous improvement compared to the more established techniques employed for fabricating similar structures.
Surface plasmons have been exploited since ancient times to produce the vivid colors observed in stained glass fabricated using metallic powders. However, many centuries passed before the underlying phenomena involved in creating the beautiful colors was identified as localized surface plasmon resonance (LSPR) in metallic nanoparticles.9 The color effects can be explained by the dissipative nature of the electron density oscillations, where the oscillation modes are excited by specific wavelengths of light, and give rise to extinction of the corresponding colors in the reflectance and transmittance spectra.2 The field of plasmonic structural colors is rapidly growing, and a more detailed description of color mapping and other recent advances was reviewed by Song et al.8 Typically, plasmonic color effects associated with metal nanoparticles are volume effects, where the colors are defined by the size and shape of the metallic nanoparticles, and the dielectric properties of the surrounding media.1 When it comes to surfaces, plasmonic colors are predominantly demonstrated with regular arrays of nanofabricated metallic structures,10–17 and there is significant interest in literature for the large scale fabrication of plasmonic color filters.18–21
Plasmonic color effects are only observed for metallic nanosized structures smaller than the excitation wavelength for LSPR, which is why such features are referred to as “sub-wavelength” structures.22,23 Hence, plasmonic color generation on solid surfaces requires the critical dimensions (CD) of the metallic nanostructures to be smaller than roughly half the excitation wavelength, i.e. effectively below ∼200 nm.17,24,25 This requirement severely limits the applicability of most lithographic techniques for fabrication of plasmonic metasurfaces. Instead, most studies are carried out using electron beam lithography to define the structures.11,12,26–31 Although electron beam lithography is indeed suitable in terms of the required resolution, as it can easily produce structures with CD below 200 nm, the throughput of the technique is extremely low. Therefore, a different approach is required for producing large area plasmonic color surfaces. Although replication roadmaps by use of nanoimprint lithography11,12 and roll-to-roll processes exist,14,32,33 large area master origination methods for the fabrication of plasmonic metasurfaces have not yet been demonstrated. When looking into the nanofabrication toolbox,34 a number of bottom-up methodologies35 such as block-copolymer nanolithography,36–41 and colloidal lithography42,43 may offer promising roadmaps for origination of large area plasmonic surfaces. In addition, depending on the light source used, interferometric lithography may also be a suitable candidate for the fabrication of large area nanoscale patterns, but this technique usually requires complicated custom made setups.44
A more direct route is to employ projection photolithography tools similar to the ones used by the semiconductor industry for the fabrication of computer chips.45–47 These tools range from standard I-line photolithography using 365 nm mercury arc lamps with resolution down to 350 nm to extreme UV (EUV) using a 13.5 nm Sn laser-produced plasma light sources featuring sub 20 nm resolution.48 EUV tools are however not available for academic research for the time being, and I-line lithography is not capable of producing structures with CD below 200 nm. To improve resolution, i.e. to decrease the feature sizes and pitches of printed patterns with acceptable tolerances, numerous resolution enhancement techniques (RET) have been introduced over the last decades in projection lithography. The range of applied RET comprise optical proximity corrections, such as the use of sheriff and assist features, lithography-based/etch-based double patterning, cross exposure, source mask optimization, and the application of alternating phase-shift masks and off-axis illumination.47,49–51 In the case of periodic patterns with pitches at the resolution limit, off-axis illumination (OAI) is the preferred RET for maximizing resolution. Hence, with OAI using e.g. a dipole source, pattern definition can be obtained with a line-grating mask theoretically having a pitch down to half the one obtained with center-axis illumination for the same tool.
In this paper, we demonstrate the fabrication of large area plasmonic surfaces by employing a innovative RET combination on a direct projection lithography deep UV (DUV) stepper tool from the ‘90s. Briefly, we supplement some of the already mentioned RETs of dipole OAI and cross exposure, with a novel approach of altering the linewidths by adding transparent assist feature (AF) lines to the mask layout. This process enables fabrication of hole and pillar patterns with different CD without changing the exposure dose and focus, and the smallest feature size generated is ∼100 nm. The combined RET is a pure exposure process, thus no post exposure processes (e.g. post exposure bake or etching) are applied between the different exposure steps, which renders complicated alignment procedures unnecessary. Table 1 shows a brief comparison between this approach and other nanopatterning methods that show promise as large area master origination methods because of their ability to produce nanostructured patterns larger than 1 cm2.
This work | Telecka et al.41 | Cataldo et al.42 | Xia et al.44 | |
---|---|---|---|---|
Nanopattern fabrication technique | Projection photolithography + RET combination | Block copolymer lithography (with reactive ion etching) | Colloidal lithography (with O2 plasma etching) | Interferometric lithography |
Resolution | 90 nm wall thickness, 159 nm pillars, 210 nm holes | Pitch ∼ 72 nm | ∼100 nm (authors claim 50 nm) | 48–91 nm (¼ of laser wavelength), 33 nm with water immersion |
Patterns | Pillars or holes | Holes or dots (pillars) | Split ring resonators | Various periodic patterns |
Pattern size | 80 × 72 mm2 (∼60 cm2) written in a few minutes, 22 × 22 mm2 for single exposure | 100 mm wafer | 1 cm2, limited by the homogeneous PMMA area fabricated using spin-coating | 1 cm2 or more |
Using the innovative RET combination for the deep UV lithography process, we demonstrate super-resolution fabrication of a color palette of structural colors, including blue, green, and red shades, merely by changing the linewidth of the AF lines on the mask. This approach allows us to exploit the DUV stepper technology for the fabrication of large area plasmonic metasurfaces, i.e. covering most of a 150 mm wafer.
The material stack consisted of a 360 nm thick layer of a positive tone resist (KRF-M230Y from JSR) on a 65 nm thick bottom anti-reflective coating (BARC, DUV-42s from Brewer Science), both spin coated onto 150 mm double side polished silicon wafers. A source stopper plate fitted with two off-axis symmetrical apertures was employed as dipole illumination source with σC = 0.6 and σR = 0.2. The exposure sequence consisted of 4 sequential exposure steps. After the first exposure step, the wafer was translated along the x-axis by a half-pitch. After the second step, both the illumination source and the mask containing the assist features were rotated by 90°. After the third step, the wafer was translated along the y-axis by a half-pitch. After the exposure sequence, a post exposure bake (60 s; 130 °C) step was included, and finally the structures were developed (60 s in AZ 726 MIF Developer from MicroChemicals). The final height of the resulting nanoholes and nanopillar structures was ∼330 nm.
10 nm thick films of either Al or Au were deposited onto the samples by directional electron-beam evaporation using the evaporator tools Temescal FC-2000 (Ferrotec) or QCL 800 (Wordentec). In all cases, the base pressure directly before metal deposition was lower than 10−6 Torr and the deposition rate was 2 Å s−1.
(1) |
Without RET, the 248 nm DUV stepper tool typically yields CD ≳ 250 nm for NA = 0.6 corresponding to k1 ∼ 0.6. Super resolution is obtained by reduction of the k1-factor in the Rayleigh criterion (eqn (1)) towards its theoretical limit of 0.25, and OAI is the best RET option.53,54 With OAI using e.g. a dipole source, pattern definition can be obtained with a line-grating mask theoretically having a pitch down to half the one obtained with center-axis illumination for the same tool. However, lens aberration effects, wafer roughness, and finite resist thickness make it practically impossible to reach the theoretical resolution limit. Here, we report a novel RET approach based on OAI double exposure and cross exposure, supplemented by the inclusion of transparent assist feature (AF) lines of various widths to the mask layout. With this approach, we obtain k1 ∼ 0.38, which is comparable to state-of-the-art in the semiconductor industry using this kind of DUV stepper tool.53
In Fig. 1, we show an overview of the process and two examples of printed plasmonic metasurfaces. Fig. 1a shows a schematic of the beam traced through the optics of the DUV tool used in this study. If the on-axis illumination source is replaced by a dipole source (two symmetrically placed sources in off-axis position), the −1st diffraction order of the beam illuminating the mask from its right side and the +1st order illuminating from the left will fall outside of the lens pupil. However, the 0th, +1st and −1st orders will pass through the lens symmetrically and form an image. In terms of Fourier optics, this corresponds to keeping +1st and −1st order spatial frequency Fourier terms to represent the periodic pattern whereas the 0th order yields a non-diffractive bias. The exposure sequence is shown in Fig. 1b. The first and second exposures are done with the same dipole illumination but with a half pitch wafer translation along the x-axis. Then both the illumination source and the mask are rotated by 90° before the third exposure. For the fourth and last exposure, the wafer is again moved by half a pitch length, but now along the y-direction. The pitch and space width of the line-grating are kept at 600 nm and 150 nm respectively, but transparent AF lines ranging from 100 to 150 nm (in steps of 10 nm) are added to the mask layout (see also ESI, Fig. S2†). In Fig. 1c, we show the resulting resist patterns after post-exposure bake and development. We see that in this example, we obtain holes with diameters of ∼210 nm and pillars with diameters of ∼159 nm for the two extreme values of AF widths, 100 nm and 150 nm respectively, arranged in a 2-dimensional 300 nm pitch array. After directional coating of the developed resist patterns with 10 nm thick Al or Au, plasmonic metasurfaces are formed, and plasmonic color generation is readily observed. This is shown in the colored inserts of the scanning electron micrographs. The shown colors are obtained from computed RGB color coordinates, which in turn are extracted from measured reflectance spectra. See also ESI, Fig. S3† for the full AF range.
In order to determine the exposure conditions required to obtain a given resist relief, we simulated cross-exposure processes with different illumination source shapes to obtain aerial images corresponding to hole and pillar patterns with quadratic symmetry and pitches of 300 nm (see ESI, Fig. S4†). An aerial image represents the optical image of the mask pattern, projected onto the image plane of the projection lens.48 Because the image is formed in air and not in a resist, the performance of the illumination tool can be simulated independently from both resist properties and subsequent post processes, i.e. post-exposure bake and development. The aerial image contrast (AIC, eqn (2)) and the normalized image log-slope (NILS, eqn (3)) were used as metrics to quantify the optical performance for the different single and cross-exposure processes:
(2) |
(3) |
AIC describes the degree of modulation in the aerial image having a maximum of 1. Here Imax and Imin are the maximum and minimum normalized light intensities. NILS is the slope of the normalized natural logarithm of the aerial image intensity function I at the edges of the target CD, whereas w is the target width. NILS is a metric for the edge definition. AIC and NILS were determined both in focus and in defocus positions to estimate the focus range inside the finite resist thickness. The simulations revealed that dipole illumination with double line cross exposure (DLCE) and application of the AF generally yields the highest values of NILS and AIC, when compared to other exposure modes of the stepper (see ESI, Fig. S5†). Here, “double line” refers to using a double linewidth of the opaque bands on the mask pattern (see Fig. 1b). This results in a mask pitch that is twice as large as the target pattern pitch, as e.g. the 600 nm pitch referred to in Fig. 1b yields a structure pitch of 300 nm when employing DLCE (see also ESI, Fig. S2†). The assist features in the DLCE process are transparent bands. Thus, a larger AF width adds more light in the image plane and increases the intensity of the areal image. This in turn decreases the pillar width in the real exposure. On the other hand, when the widths of the AF are increased, a lower dose is required to print a particular pattern with the same dimensions. However, simulations of NILS and AIC show that the width of the AF does not affect the quality of the areal image. This is shown in ESI, Fig. S5,† a category plot, which compares simulations of NILS and AIC for exposures with different assist feature sizes and source shapes. We see that, while dipole illumination is not beneficial for large pitches, the introduction of AF improves dipole illumination with respect to the other illumination modes. It is remarkable that the width of the AF does not affect the areal image parameters significantly. On the other hand, compared to the results of the cross-exposure process (Fig. S4 and S5†) single exposure processes with similar source shapes are characterized by much lower NILS and AIC values (not shown here).
In Fig. 2, we document the plasmonic color generation from the fabricated surfaces coated with thin metal layers (10 nm of Au or Al). The colors of the metallized resist patterns are shown in the form of color palettes (Fig. 2b). By varying the widths of AF on the mask, we see a transition from hole-arrays to pillar-arrays in the resist after development (see also ESI, Fig. S3†). Apart from the widths of AF, this transition depends on the focus offsets employed during exposure. The color palettes obtained with different focus offsets are included to demonstrate the color stability upon focus variations that could occur during the exposure process due to e.g. wafer roughness, lens aberrations and inhomogeneous focus offsets at different exposure positions on the wafer. The 0 μm defocus is related to the “best focus”, i.e. a focal point in the resist at a position approximately a third of the resist thickness below the surface of the resist. For a dose of 65 J m−2, employed for the structures shown in Fig. 1c and 2, this transition happens at AF = 120 nm, so that assist features below 120 nm yield hole array surfaces, whereas assist features widths of 120 nm and above yield pillar array surfaces. Whilst a defocus smaller than 0.2 μm has very little influence on the observed plasmonic color generation, a defocus of 0.2–0.4 μm has a more noticeable effect.
In order to ensure that colors do not depend on the light source used, we obtain the colors in Fig. 2b in the form of RGB color coordinates from the measured reflectance spectra that are normalized with the spectrum from the light source. The ESI† shows the CIE diagrams with the measured colors in Fig. S6,† and the RGB color coordinates corresponding to the palettes displayed in Fig. 2c in Table S1.† We see that whereas we do not obtain pure primary red, green and blue colors, we obtain mixed colors where one primary value typically dominate. For metallization of the resist structures with 10 nm Al, we see a dominant peak in the reflectance spectrum developing from the green part of the spectrum around 500 nm, when AF widths are increased from 100 nm to 150 nm. Hence, for the bright green color, obtained with Al coating of the developed resist (at 0 μm defocus) with AF = 100 nm, we get (RGB) = (0, 0.616, 0.467), whereas the corresponding most blue color for AF = 130 nm yields (RGB) = (0, 0.314, 0.470), and for the most red color for AF = 150 nm we have (RGB) = (0.387, 0.101, 0.230). Whereas the Al metallization exhibits the most pure green colors, metallization with Au yield the most pronounced red colors with e.g. AF = 140 nm yielding (RGB) = (0.542, 0.368, 0.228) for in focus exposure. The Au coated samples however also tend to have high contents of green, resulting in brownish shades.
It can be seen from Fig. 2d that (i) for the Au structures, the simulated reflectance spectra agree well with those obtained experimentally. (ii) For the Al structures, the simulated reflectance spectra have similar spectral fingerprints compared to the measured ones, but there is a red shift of ∼60 nm. This red shift may be caused by a difference between the complex refractive index values as a function of wavelength for Al used in the simulation and that of the real structures, which could be attributed to the thickness of Al used in this study. The influence of the metal layer thickness on the refractive index has previously been measured for Au and Ag.55 A similar behavior is expected for thin Al layers, but, to the best of our knowledge, the complex refractive index of 10 nm thick Al layers cannot be found in literature. In addition to the thickness-related refractive index issue, minor differences between the simulated and the measured reflectance spectra may appear due to the following factors: (i) in the simulation, the geometry of the structures is identical in every period. However, structural variations across periods may exist in the real structures. (ii) In the simulation, the nanopillars and the nanoholes are perfect cylinders, which is slightly different from the real structures. (iii) In the simulation, all reflected and scattered light contribute to the reflectance, while only a portion of the scattered light contributes to the measured reflectance spectra due to the collection angle limitation in the optical reflectance measurement setup.
We also investigated the influence of different linear polarizations at normal incidence on the plasmonic color generation for Al coated structures. The results are shown in Fig. 3. The color response was measured experimentally for linear polarizations oriented at angles of 0°, 45° and 90° relative to the structure arrays, and the corresponding CIE diagrams are shown. We determined that the influence of the linear polarization rotation is very limited, especially for the features printed with a focus offset of 0 μm. Furthermore, the finite element model simulations shown in Fig. S7† confirm that the filter properties of the Au metasurfaces at normal incidence also are nearly independent of the linear polarization direction. This is attributed to: (i) the symmetry of the structure, resulting in the same LSPR mode being excited for a single unit structure under different linear polarization directions. (ii) The relatively large pitch of 300 nm, with which the LSPR couplings among the periods become negligible. This is also reflected in the field distribution in Fig. S7† by the very weak field enhancement near adjacent periods.
Fig. 3 CIE diagrams for the Al metasurfaces obtained with linear polarizations of (a) 0°, (b) 45°, and (c) 90° measured relatively to the surface structure array. |
The similar spectral fingerprints of the simulated and the measured reflectance spectra validate our understanding of the localized surface plasmon resonance (LSPR) modes of the fabricated structures. To demonstrate, Fig. 4 shows the calculated cross-sectional distributions of the electric field enhancement, for one period of Au nanopillar (AF = 150 nm) and Au nanohole (AF = 100 nm) arrays at excitation wavelengths corresponding to the reflectance maxima and minima in the simulated spectra shown in Fig. 2d. It can be seen from Fig. 4 that the main LSPR mode for the Au nanopillar and the Au nanohole array, is the Au nanodisc and the Au nanohole mode, respectively.12 The experimental measurements and simulations described herein are for normal incidence. It is likely that non-normal incidence angles would introduce additional LSPRs oscillating perpendicular to the structural surface and could result in differences in the observed plasmonic color generation. However, this was not investigated as part of the current study.
Fig. 4 Simulated cross-sectional distributions of the electric field enhancement for one period of the (a) Au nanopillar array (AF = 150 nm) and (b) Au nanohole array (AF = 100 nm) for excitation wavelengths corresponding to the reflectance maxima and minima from the simulated spectra shown in Fig. 2d. |
Combining the field distributions shown in Fig. 4 with the simulated reflectance spectra from Fig. 2d, it can be concluded that the LSPR wavelengths result in the reflectance minima, and that the LSPR wavelengths can be tuned by altering the structural feature size. In addition, Fig. 4 reveals that the strengths of the LSPR relate to the interference between the incident and the reflected light. Resonances are achieved when the Au nanodiscs (AF = 150 nm) and the Au nanoholes (AF = 100 nm) are near the constructive interference patterns of the excitation fields. This means that by changing the aspect ratio of the structures, further tuning of the reflectance, i.e., color, could be achieved.
Kumar et al. reported a pixel size as small as 250 nm, i.e. close to the diffraction limit of optical microscopy.11 However, such small pixel sizes are not needed for color generation. Fig. S3† contains SEM images showing resist patterns for different hole and pillar arrays with different critical dimensions, which are used in the fabrication of plasmonic color arrays. The pitch of the structures – in horizontal and vertical direction – is 300 nm (see Fig. S2†). With an optical microscope having its resolution in the micrometer range, an array containing at least 10 structures in each direction (or approx. 3 × 3 μm2) is needed to observe a distinct color reflection. Hence, the proposed technology will allow in excess of 71 million pixels per square inch and 6 colors. There is also an issue regarding alignment accuracy: In one pixel field the maximum alignment offset between the first two exposures and the two exposures with the 90° rotated reticle is ∼60 nm. Inside the individual color pixel this misalignment does not affect the efficiency of the array, however on the edges between different pixels there will be a region of maximum 60 nm with color blur. This is, however, not observable in an optical microscope. In addition, printing multicolor high resolution images over entire wafers with the technique presented here will be possible, but will require a computer algorithm to design the two reticles.
Whereas primary colors were not demonstrated in this paper, it is worth noting that a number of factors influence the reflectance and therefore the color appearance, such as the sizes of the AF employed and the metal chosen for coating. Furthermore, simulation results indicate that further tuning can be achieved by altering the aspect ratio of the structures. Overall, our approach shows excellent flexibility in terms of selecting the color appearance for relatively large plasmonic metasurfaces.
LSPR | Localized surface plasmon resonance |
CD | Critical dimensions |
EUV | Extreme UV |
RET | Resolution enhancement technique |
OAI | Off-axis illumination |
DUV | Deep UV |
NA | Numerical aperture |
DRIE | Deep reactive ion etching |
DREM | Deposit, remove, etch, multistep |
BARC | Bottom anti-reflective coating |
SEM | Scanning electron microscopy |
EHT | Electron accelerating voltage |
FEM | Finite element method |
AF | Assist features |
AIC | Aerial image contrast |
NILS | Normalized image log-slope |
DLCE | Double line cross exposure |
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0na00934b |
This journal is © The Royal Society of Chemistry 2021 |