Singapore Researchers Deploy Machine Learning and Photonic Metasurfaces to Revolutionize Light Confinement for Future Optoelectronics

Singapore researchers utilize machine learning and metasurfaces to trap light at the nanoscale, paving the way for more efficient sensors and microLEDs.

By: AXL Media

Published: Apr 14, 2026, 11:53 AM EDT

Source: Information for this report was sourced from EurekAlert

Engineering the Future of On-Chip Light Control

A research team at the Agency for Science, Technology and Research, Singapore, has detailed a transformative approach to light manipulation that could redefine the efficiency of modern electronic devices. By focusing on metasurfaces—ultrathin layers composed of nanoscale "meta-atoms"—the scientists are moving toward a flat, chip-compatible platform that offers unprecedented control over photon behavior. This breakthrough centers on the application of photonic bound states in the continuum, a physical phenomenon that allows light to be trapped in open systems, creating a high-interaction environment between light and active media that was previously difficult to sustain at such small scales.

The Physics of Mirrorless Light Trapping

Traditional optical cavities have historically relied on light reflecting between two mirrors, a mechanism that becomes increasingly difficult to scale as devices shrink. In contrast, bound states in the continuum trap light through destructive interference between light waves, allowing photons to remain confined even within open structures. While theoretical models suggest these states can trap light indefinitely, the A*STAR team, led by Dr. Son Tung Ha, has focused on "quasi-BICs." These practical iterations retain exceptionally strong confinement despite minor leakage, positioning them as the most viable optical resonators for the next generation of compact, energy-efficient optoelectronics.

Harnessing Machine Learning for Nanoscale Design

As the functional requirements for modern sensors and cameras grow more sophisticated, the design of these nanostructures has surpassed the capabilities of traditional symmetric modeling. The research highlights a shift toward advanced design strategies, including machine learning and inverse-design approaches, to navigate the complexities of light control. According to Dr. Ha, these automated tools are becoming essential as metasurfaces evolve to handle multifaceted tasks. By utilizing these computational methods, researchers can now identify optimal material platforms and geometries that were previously too complex to conceive through manual engineering.