Massive scale-up of two-photon lithography enabled by fundamental science
Two-photon lithography (TPL) is an additive manufacturing technique for fabrication of arbitrarily complex 3D structures with submicron features. Although TPL uses light to polymerize material, it can print sub-diffraction 3D volumetric pixels smaller than 200 nm. This capability makes it particularly attractive for printing of optical and mechanical metamaterials, photonic crystals, micromachines, and micro-optics. However, the slow speed of the process (~0.01 mm3/hr) makes it impractical for many applications. Past scale-up attempts have failed to achieve printing of fine submicron features. We have increased the processing rate by more than a thousand times by using time-domain focusing of femtosecond light. We demonstrated this by printing, within single-digit millisecond time scales, nanowires with widths smaller than 175 nanometers over an area one million times larger than the cross-sectional area. Our current work in this area includes elucidating the process mechanisms to enable deterministic printing of target 3D structures and expanding the material palette beyond polymers.
Publication: Saha et al., Science, 2019. https://doi.org/10.1126/science.aax8760
Affordability of nanoscale additive manufacturing
The high cost of nanoscale additive manufacturing equipment (relative to commercial hobbyist 3D printers) is a significant barrier to its widespread adoption in real-world applications. A recurring theme in nanoscale patterning is the tradeoff that exists between precision (i.e., a surrogate for performance) and equipment cost. We have demonstrated how one may break this tradeoff by enabling multi-material nanoscale 3D printing capability using precise yet affordable equipment. Our current work in this area includes exploring novel light-matter interactions to generate low-cost nanoscale additive manufacturing capability.
Publication: Saha et al., Precision Engineering, 2018. https://doi.org/10.1016/j.precisioneng.2018.05.009
Machine learning for manufacturing scalability
The scalability of several emerging advanced manufacturing processes is limited by the slow, expensive, and high-skills manual process steps. It is often not possible to automate these steps through coded instruction sets (i.e., traditional algorithms) because these process steps rely on learned skills (or ‘process intuition’) of the human operator. Scalable implementation of these process steps may be achieved via machine learning algorithms. We have recently demonstrated how machine learning models can learn to differentiate and classify defective vs non-defective parts printed by TPL. Our current work in this area involves applying machine learning to automate the interpretation and classification of metrology data for micro and nanoscale additive manufacturing and in inverse design of 3D printed structures.
Publication: Lee et al., Additive Manufacturing, 2020. https://doi.org/10.1016/j.addma.2020.101444
Process predictability through nanoscale metrology
Offline measurements of the final printed geometry are often the only measure of process outcome that can be used to verify predictive process models of nanoscale additive manufacturing. This limits our process knowledge and forces us to make assumptions that may not be accurate. Nanoscale metrology can significantly expand our process knowledge and predictive capability by providing previously inaccessible data on process conditions. For example, we have demonstrated how printing under low and high-speeds can generate nanowires of the same geometry but differing stiffness and strength – an observation that cannot be estimated through geometrical measurements alone. We have also used nanoscale X-ray computed tomography (CT) to clarify that thermal damage in two-photon lithography is not a stochastic defect but can be accurately predicted and controlled. Our current work in this field involves generating in-situ metrology capabilities to monitor process conditions during nanoscale polymer and metal printing.
Publication: Ladner et al., RSC Advances, 2019. https://doi.org/10.1039/C9RA02350J