Large Area Patterning of Nanoarchitected Metamaterials via Metasurface Mask Interference Lithography

Author: Thomas Tran

Mentors: Julia Greer, Ryan Ng, Phillippe Pearson

Editor: Jonathan Chan

Introduction 

Materials architected on the nanoscale exhibit nonintuitive mechanical properties unlike those observed in bulk materials. For example, Professor Julia Greer’s work in architecting nanolattice structures from alumina has defied conventional knowledge that ceramics must be brittle [1]. In addition to enabling elastic recoverability in ceramics, these nanolattices demonstrate that tensile strength and density can be decoupled and illustrate the possibility of manufacturing nanoarchitected metamaterials with remarkable densities and mechanical properties. Currently, however, large-scale application of these materials is limited by the lack of fabrication methods that can pattern large samples with sufficient resolution. For this reason, there is great interest in devising new, scalable processes that can generate large samples whose mechanical properties improve upon those of traditional bulk materials.

3-D periodically patterned nanostructures with sub-micron features are of significant interest because of their applications in photonics, phononics, and nanofluidics [2][3][4]. Specifically, they can be used as photonic or phononic crystals with sub-wavelength sized features that can be tuned to engineer the propagation of electromagnetic or acoustic waves, respectively. In nanofluidics, they can be used to control the flow of fluids. In this study, we are interested in these structures’ low relative density and improved energy absorption properties arising from their nanoscale trusses because they have potential applications as lightweight, highly impact resistant materials. However, a more scalable fabrication method is required to produce these structures because current techniques cannot manufacture them efficiently in bulk quantities. 

Current methods for synthesizing these periodic nanostructures include various types of lithography, including two-photon lithography and interference lithography. Lithography generally involves the selective irradiation of a photoresist material—a photosensitive polymer—thereby leaving a pattern. The lithography process can vary depending on the illumination sources, photoresist materials, and phase-shifting masks used. For 3-D nanolattice fabrication, two-photon lithography is typically favored because its high spatial resolution allows for the creation of complex architectures [2][5]; however, this approach’s serial spot-by-spot approach is time-consuming and limits its ability to pattern large areas. 

In contrast, interference lithography uses a parallel approach to rapidly fabricate cheap 3-D structures on the sub-micron scale [4]. In a holographic approach to interference lithography, multiple coherent laser beams can interfere with each other and form periodic intensity distributions within the resist material, thus exposing a pattern in a single exposure. 1-D, 2-D, and 3-D patterns can be generated using 2, 3, and 4 beams respectively [4]. Additionally, the modification of laser configurations can result in customizable periodic patterns with body-centered cubic (bcc), face-centered cubic (fcc), gyroid, or diamond cubic symmetries [2]. However, while interference lithography serves as an efficient method for patterning large areas of photoresist in a single exposure step, the holographic approach can be difficult to implement as one must carefully modulate beam directions, polarizations, and phase shifts of multiple interfering beams. Controlling these parameters makes holographic interference lithography difficult to scale up with a raster patterning setup, since this setup must additionally scan the beams across photoresist and introduce external vibrations. 

To minimize these difficulties, the use of masks in interference lithography can replicate interference patterns without the need for multiple beams. Jeon et. al have used conformal phase masks made from conventional diffractive optics to form 3-D nanostructures [5][6]. Likewise, the development of optical metasurfaces—2-D arrays of subwavelength scatterers—allows for intensity distributions of light to be engineered based on the scatterer design and placements. Metasurface masks tend to be thin and planar, with the potential to be mass-produced at a low cost using standard micro- and nanofabrication processes [7]. Like conformal phase masks, metasurface masks in interference lithography can be used to reliably generate 3-D nanostructures in a single exposure step without the need for a complex four-beam setup [6][7][8]. In this study, we propose exploiting the advantages of metasurface mask interference lithography to create a scalable process for fabricating 3-D periodic nanostructures. Using a precise translational sample stage, we combine adjacent single exposures to form one large raster pattern of exposures to achieve our desired nanostructures.…

Materials and Methods

Our samples consisted of a stack of fused silica substrate and two layers of SU-8, a high-contrast negative photoresist commonly used in lithography. The first layer of SU-8 was a fully cured adhesion layer between the fused silica and the nanoarchitected SU-8 layer above.  

To implement our lithography technique, we spin coated sensitized SU-8 at 1000 rpm onto a fused silica substrate and adhesion layer.  Following the removal of edge beads with a razor, the sample was baked at 65°C for 10 minutes and 95°C for 20 minutes to remove solvent. Next, the sample was exposed to a 532 nm laser beam passed through a metasurface mask, thus generating the photoacids that allowed the SU-8 monomers to crosslink upon heating. We then heated the samples at 65°C for 3.5 min before cooling them and allowing the exposed regions of the sensitized resist to crosslink. The samples were placed into a propylene glycol methyl ether acetate (PGMEA) bath for two hours and transferred to an isopropanol bath for 30 min to dissolve away unexposed regions. Finally, critical point drying was used to complete the development process and preserve the SU-8 nanostructures.

For simple exposures, the beam setup consisted of a continuous 532 nm laser with adjustable power, a polarizer, and a flat-top beam shaper. The flat-top beam shaper converted the Gaussian intensity distribution of the beam to a uniform distribution while expanding the beam diameter from approximately 1 mm to 1 cm. As a result, this beam setup produced structures approximately 1 cm² in size. Exposing just one spot on the samples yielded consistent patterns laterally, resolving a periodic pattern with a pitch, or spatial period, of ~1 μm. In the axial direction, the pattern’s pitch was elongated to ~2 μm. The thicknesses of fabricated samples were limited to 20 μm. On this scale, we employed two raster patterning methods: tiling and continuous exposure. 

In tiling exposure, a large area of sensitized photoresist was patterned by performing multiple, discrete exposures adjacent to each other (Fig. 1). Each exposure area formed a voxel of a nanostructured area, and multiple exposure areas were stitched together to form a larger sheet. Exposures were carried out using a beam shutter and translational sample stage, in which the metasurface mask was coupled with the laser as it moved across the sample stack between exposures. While this method could potentially fabricate very large samples, it led to heterogeneities within each exposure area that were dependent on the beam profile (Fig. 4a). Additionally, overlapped exposure areas (stitch areas) resulted in overexposure. Consequently, following development, these stitch areas became solid blocks of resist and did not contain the desired nanoarchitecture.  

In continuous scanning, large areas of photoresist were patterned by scanning the laser beam across the sample surface at a constant velocity (Fig. 2). In this case, the metasurface mask was coupled with the moving sample stack, instead of the laser. Heterogeneities that were previously present in the beam profile were reduced, leading to a more uniform sample morphology [8]. However, because the metasurface was coupled with the sample stack, the maximum patternable area with this method was limited by the size of the metasurface mask (currently around 1 cm²).

For both tiling and continuous scanning exposures, a translational X-Y sample stage (Physik Instrumente V-5517B Precision Linear Stage) and a motorized shutter (Thorlabs SH1 Optical Beam Shutter) were used to control the precise movement and timing of exposures. The shutter control and stage movement were coordinated through a custom MATLAB graphical user interface (GUI) developed for large area exposures. The MATLAB interface took advantage of the stage’s 2 nm motion resolution and automated entire exposures based on the tiling or continuous scanning methods (Fig. 3). For the tiling method, the program opened and closed the shutter for a specified exposure time for multiple static exposures. For the continuous scanning method, the program opened and closed the shutter once per row and moved the stage at a constant velocity as the shutter was open.

Several single exposures were performed to observe the consistency of the patterns formed with an air gap and a square aperture (0.7 x 0.7 cm²). Then, we patterned three samples using various patterning methods and beam profiles to analyze their effect on the samples’ nanoarchitected morphologies. For the first two samples, we performed one simple continuous scan exposure and one simple tiling exposure using the square aperture and flat-top beam shaper. The continuously scanned sample was fabricated through the dynamic exposure of one line, and the tiled sample was fabricated by overlapping two static exposures. Last, we observed the effects of the beam intensity profile on pattern consistency by fabricating another continuously scanned sample without the use of the flat-top beam shaper. A beam with Gaussian intensity distribution was scanned without the use of any aperture to avoid aperture diffraction effects.

Results and Discussion

Individual spot exposures using the metasurface mask interference lithography setup indicated that the mount-aperture could successfully pattern a square area without patterning outside the target area. However, these samples contained ripple-like features throughout the cross-sections of the nanostructures parallel to the aperture (Fig. 4a). This may have been due to diffraction effects from the square aperture, in addition to heterogeneities present in the beam’s intensity distribution despite the use of a flat-top beam shaper (Fig. 4b). These beam heterogeneities caused volumes of photoresist to become unevenly exposed, and as a result, created areas of extreme distortion and structural instability. Furthermore, the shape of unit cells as well as cell wall thicknesses were not uniform and differed from samples patterned without an SU-8-metasurface air gap. We performed a simple tiling experiment to observe the effects of overlapping exposures. 

The first tiling exposure was performed by exposing two adjacent areas with a large region of overlap.  The non-overlapped area received an appropriate dosage of light and exhibited a somewhat consistent pattern in local areas (Fig. 5a). However, heterogeneities that were observed in single exposures continued to appear in tiled samples. The overlapped area instead produced a block of solid resist due to receiving a higher dosage of 532 nm light. Thus, the tiling method introduced a hierarchical structure of nanoarchitected regions with solid boundaries. However, further testing is needed to understand how such interfaces affect the samples’ mechanical properties. Future improvements to the tiling method might also involve adjusting the widths of the solid boundaries using the high precision of the sample stage. 

The first continuous scanning exposure was performed by scanning a line at constant velocity across the sensitized resist. A cross section of the resulting figure is shown in Fig. 5b, which demonstrates consistent patterning in all three axes. It had a period of approximately 1 μm in the lateral directions, while the axial direction was more elongated with a period of ~4 μm. These periods were comparable to those of statically exposed samples, where the period was approximately 1 μm in the lateral directions and 2 μm in the axial direction. Differences in axial period were also observed despite using the same metasurface mask. Additionally, unit cells were taller in the middle of the continuously scanned sample, an effect not observed in previous single-exposure or tiled samples. The patterning, however, was much more consistent through the cross section and top surface than the samples produced with static exposures (including tiling). More areas were appropriately exposed, presumably due to the movement of the beam averaging out heterogeneities in the beam profile that would translate to areas of underexposed resist. In comparison to the tiling method, the continuous scanning method did not produce overexposed regions within the continuously scanned rows, minimizing defects. 

From these preliminary results, we decided that the continuous scanning method for large area exposures was favorable over tiling primarily due to its mitigation of defects in samples. However, the tiling method was still useful in terms of scalability, as it could fabricate nanoarchitected sheets up to 23 x 23 cm² in size (limited by the travel range of the translational sample stage). By comparison, the continuous scanning method could only fabricate samples of sizes no larger than that of the metasurface mask (0.7 x 0.7 cm² at the time of experiment). The shortcoming of the tiling approach was that individual spot exposures that make up the raster pattern were not always consistently patterned, leaving patches of underexposed areas as seen in Fig. 5a.  

These pattern inconsistencies may be due to the use of the flat-top beam shaper as well as the rectangular aperture. The use of the flat-top beam shaper did not create an ideal, uniform intensity profile, as propagation of the beam distorted its intensity (Fig. 4b). Additionally, the rectangular aperture introduced further diffraction effects and heterogeneities in the beam’s intensity profile. Thus, if the beam shaper and aperture were removed from the setup and a Gaussian beam were used instead, we hypothesized that the beam profile would be more predictable and would lead to more consistent results. 

To test this approach, we fabricated another sample using a Gaussian beam and the continuous scan method with overlap significant enough such that that sum of the Gaussian distributions were relatively uniform in comparison to the experimental flat-top distribution. Four rows were continuously scanned at constant velocity and produced samples that were even more uniform than those made using a flat-top beam (Fig. 6). While some heterogeneities still existed, the Gaussian beam led to the most consistently patterned samples yet. The thicknesses were consistent, the presence of improperly exposed areas was minimal (Fig. 6a, b) and the cross section retained its periodic features (Fig. 6c). However, a downside to the approach was that the beam size was reduced from a diameter of 1 cm to 2.2 mm; as such, increased processing times were needed to minimize defects and promote uniformity. 

The success of the continuous scan using a Gaussian beam demonstrated the possibility of fabricating a defect-free sample with the tiling method as well. Since the intensities are continuously distributed, it is possible to overlap them such that overexposed stitch areas are minimized. By avoiding the use of a rectangular beam or a flat-top profile, overexposed areas would exhibit a gradual as opposed to an abrupt transition between solid and patterned resist. This would be favorable for fabricating large, uniform nanoarchitected surfaces that retain low relative density and mitigate stress concentration upon loading.

Conclusions

While two-photon lithography is commonly used for making 3-D nanoarchitected materials, the approach is time-consuming and cannot efficiently produce macroscale objects. In this study, we aimed to improve upon these limitations by demonstrating that metasurface mask interference lithography can be used to produce nanostructures on the order of 1 cm² in area and 20 μm in thickness in one step. We successfully scaled up these materials by stitching together voxels of SU-8 photoresist using a translational sample stage and a motorized shutter controlled by automated software. Three approaches to raster patterning were tested: tiling with an expanded flat-top beam; continuous scanning with an expanded flat-top beam; and continuous scanning with a Gaussian beam.  

Initial experiments for scaling up the exposure process showed differences between continuous scanning and tiling methods for large area exposures. While continuous scanning exposed samples uniformly, tiled samples possessed overexposed stitch areas and intraregional heterogeneities. These heterogeneities were minimized using a Gaussian beam setup (instead of a flat-top beam setup), which produced more predictable beam profiles and exposures. Our work illustrates the promise of using metasurface mask interference lithography to produce larger scale nanoarchitected materials with potentially remarkable mechanical behaviors.

Future work will focus on characterizing how patterning defects affect the mechanical properties of nanoarchitected samples and determining the optimal patterning methods for high-impact applications. We also hope to improve our lithography setup by modifying the photoresist pattern, thickness, and chemistry. Specifically, we hope to address the limitations of spin coating photoresist (which limits the thicknesses of the nanostructure sheets to approximately 100 μm) and develop post-processing techniques to allow non-polymeric materials to be patterned; for example, using pyrolysis to produce carbonized structures.

Acknowledgements

I would like to thank Professor Julia Greer, Ryan Ng, Phillippe Pearson, and members of the Greer Group at Caltech for their mentorship. I would also like to thank Professor Andrei Faraon, Farzaneh Afshinmanesh, Andrew Friedman, and Luizetta Navrazhnykh from the Scalability Team for their guidance on this project. Finally, I am grateful for financial support from nFugue and Caltech’s Student-Faculty Programs Office.

References

1. L.R. Meza, S. Das, J.R. Greer, Strong, lightweight, and recoverable three-dimensional ceramic nanolattices. Science. 345(6202), 1322-1326 (2014). doi: 10.1126/science.1255908
2. A.d. Campo and C. Greiner, SU-8: a photoresist for high-aspect-ratio and 3D submicron lithography. Journal of Micromechanics and Microengineering. 17(6), (2007). 
3. M. Campbell, D. N. Sharp, M. T. Harrison, R. G. Denning, and A. J. Turberfield. Fabrication of photonic crystals for the visible spectrum by holographic lithography. Nature. 404, 53-56, (2000). doi:10.1038/35003523
4. J. Jang et. al, 3D Micro- and Nanostructures via Interference Lithography. Adv. Funct. Mater. 17, 3027-3041, (2007). doi:10.1002/adfm.200700140
5. S. Jeon et. al, Fabricating complex three-dimensional nanostructures with high-resolution conformable phase masks. PNAS. 101(34), 12428-12433, (2004). doi:10.1073/pnas.0403048101
6. S. Jeon, V. Malyarchuk, J.A. Rogers, and G.P. Wiederrecht, Fabricating three dimensional nanostructures using two photon lithography in a single exposure step. Optics Express. 14(6), 2300-2308, (2006). doi:10.1364/OE.14.002300
7. S.M. Kamali, E. Arbabi, A. Arbabi, and A. Faraon, A review of dielectric optical metasurfaces for wavefront control. Nanophotonics. 7(6), (2018). doi:10.1515/nanoph-2017-0129
8. L. Yuan and P. R. Herman. Laser Scanning Holographic Lithography for Flexible 3D Fabrication of Multi-Scale Integrated Nano-structures and Optical Biosensors. Scientific Reports. 6:22294, (2016). doi:10.1038/srep22294

…