Glass nanostructures reflect nearly all visible light, challenging assumptions

Summary

Nanoscale 3D printing of glass. Credit: Science Advances (2025). DOI: 10.1126/sciadv.adv0267 A research team led by SUTD has created nanoscale glass structures with near-perfect reflectance, overturning long-held assumptions about what low-index materials can do in photonics. For decades, glass has been a reliable workhorse of optical systems, valued for its transparency and stability. But when it comes to manipulating light at the nanoscale, especially for high-performance optical devices, glass has traditionally taken a backseat to higher refractive index materials. Now, a research team led by Professor Joel Yang from the Singapore University of Technology and Design (SUTD) is reshaping this narrative. With findings published in Science Advances, the team has developed a new method to 3D-print glass structures with nanoscale precision and achieve nearly 100% reflectance in the visible spectrum. This level of performance is rare for low-refractive-index materials like silica, and it opens up a broader role for glass in nanophotonics, including in wearable optics, integrated displays, and sensors. The researchers' breakthrough is enabled by a new material called Glass-Nano: a photocurable resin made by blending silicon-containing molecules with other light-sensitive organic compounds. Unlike conventional approaches that rely on silica nanoparticles—often resulting in grainy, low-resolution structures—Glass-Nano cures smoothly and contracts uniformly during heating, transforming into clear, robust glass. When printed using two-photon lithography, these polymer structures shrink during sintering at 650°C, preserving their form while achieving nanoscale features as small as 260 nanometers. "Instead of starting with silica particles, we worked with silicon-bearing molecules in the resin formulation," explained Prof Yang. "This resin enables us to build up nanostructures with much finer detail and smoother surfaces than was previously possible. We then convert them into ...