Compact and lightweight meta-surfaces – which use specially designed and patterned nanostructures on a flat surface for focusing, shaping and light control – are promising technologies for wearable applications, especially virtual and augmented reality systems. Today, research teams are carefully developing a specific image of nanostructures on the surface to achieve the desired lens function, whether it be the separation of nanoscale features while creating multiple images that perceive depth or focus light regardless of polarization.
If metalens is to be used commercially in AR and VR systems, it will need to be significantly increased, which means that the number of teachers will be measured in billions. How can researchers develop something so complex? That’s where artificial intelligence comes in.
In a recent paper published in Art The nature of communicationA team of researchers from the Harvard School of Engineering and Applied Sciences John A. Paulson (SEAS) and the Massachusetts Institute of Technology (MIT) described a new method of designing large-scale metasurfaces that uses machine intelligence techniques to automatically create structures. .
“This paper lays the foundation and approach to design that can impact many real-world devices,” said Federico Capas, professor of applied physics at Robert L. Wallace and senior researcher in electrical engineering at SEAS and senior author of the paper. “Our techniques will create new metasurfaces that can affect virtual or augmented reality, cars and machine vision for systems and satellites.”
So far, researchers need years of knowledge and experience in this field to develop a metasurface.
“We were guided by design based on intuition, relying heavily on our training in physics, which was limited in the number of parameters that can be considered simultaneously, limited, like us, the ability of human working memory,” said Zhaoi Li, a scientist. a SEAS employee and one of the lead authors of the article.
To overcome these limitations, the team taught a computer program the physics of metasurface design. The program uses the basics of physics to automatically create meta-surface designs while simultaneously developing millions and billions of parameters.
This is a reverse design process, which means that researchers start with the desired function of metals – such as a lens that can correct chromatic aberration – and the program finds the best geometries to achieve this goal using its computational algorithms.
“Allowing a computer to make decisions is inherently scary, but we’ve demonstrated that our program can act as a compass, pointing the way to optimal design,” said Rafael Pesturi, a doctoral student at the Massachusetts Institute of Technology and one of the lead authors. “Moreover, the whole design process takes less than a day using a single-processor laptop, compared to the previous approach, which would have taken months to simulate a single 1cm diameter meta-surface running in the visible light spectrum.”
“This is an order of magnitude increase in the scale of reverse design for nanostructured photonic devices, generating devices with tens of thousands of wavelengths in diameter compared to hundreds in previous work, and this opens up new classes of applications for computational discoveries,” said Stephen J. Johnson is a professor of applied mathematics and physics at the Massachusetts Institute of Technology and a co-correspondent for the article.
Based on the new approach, the research team developed and produced a centimeter-scale, non-polarizing insensitive, RGB-achromatic meta-eyepiece for the virtual reality platform (VR).
“Our presented VR platform is based on meta-glasses and micro-SRS with back laser illumination, which offers many desirable features, including compactness, light weight, high resolution, wide color gamut and more,” Lee said. “We believe that the metasurface, a form of flat optics, opens a new path to rebuilding the future of VR.”
Co-authors of the study are Jun-Su Pak and Yao-Wei Huang. It was partially supported by the Agency for Advanced Defense Research Projects (grant № HR00111810001) and AFOSR (grant № FA9550-21-1-0312). This work was partially performed at the Center for Nanoscale Systems (CNS), a member of the National Coordinated Infrastructure of Nanotechnology (NNCI), supported by the National Science Foundation under the NSF прэ Award. 1541959