Metasurfaces for Compact and Spectrally Selective AR Systems

We are excited to share our recent work published in Nano Letters, “Two-Dimensional Topology Optimized Nonlocal Metasurfaces for Augmented Reality.” In this study, we combine nonlocal metasurfaces with two-dimensional topology optimization to create an ultrathin optical device that can selectively manipulate red, green, and blue (RGB) light within a single planar nanostructure. We further integrate the device into a free-space augmented reality (AR) system to demonstrate its potential as a compact optical combiner.

Why does AR need new optics?

AR glasses need to do two things at the same time: allow us to see the real world and deliver virtual images from a display into our eyes. Conventional systems often rely on multiple lenses, mirrors, or waveguide components to accomplish these tasks, adding size and complexity while introducing challenges such as optical loss, color crosstalk, and unwanted display-light leakage.

Metasurfaces offer a different approach. By engineering nanostructures much smaller than the wavelength of light, optical functions that traditionally require bulky components can potentially be integrated into a flat, ultrathin device.

But we asked a further question: Can we control not only where light goes, but also precisely which colors of light are manipulated?

When nanostructures work together

This is where nonlocal metasurfaces become particularly interesting.

In a conventional local metasurface, each nanostructure can be thought of as a tiny optical element that largely responds on its own. In a nonlocal metasurface, however, the optical response emerges collectively from interactions across many neighboring structures. These collective resonances can produce extremely narrow spectral responses, allowing the device to strongly interact with selected wavelengths while rejecting nearby colors.

In our work, we use a resonant waveguide grating (RWG) as the nonlocal photonic platform and introduce two-dimensional topology optimization to search for the optimal nanostructure geometry. Expanding the design space from one to two lateral dimensions provides additional freedom to independently tailor different resonances, retaining the desired RGB channels while suppressing unwanted spectral responses.

Letting optimization discover the structure

One striking outcome of topology optimization is that the final structures no longer resemble simple, regularly shaped gratings. Instead, the algorithm discovers freeform nanostructures tailored to the desired optical response.

By changing the optimization objective, the same framework can selectively preserve particular resonances, suppress unwanted wavelengths, or simultaneously control multiple RGB channels. Our fabricated devices experimentally confirmed the predicted narrowband responses, efficient diffraction, and strong wavelength selectivity.

From nanostructures to an AR display

Rather than stopping at simulations and spectral measurements, we integrated the fabricated metasurface into a free-space AR prototype.

Here, the metasurface acts as an optical combiner: ambient light passes through the device so that the wearer can see the real world, while selected RGB display light is redirected toward the viewer. Our experiments demonstrated vivid virtual images with high color purity while suppressing unwanted spectral leakage.

This distinction is important for practical AR. An ideal optical combiner must do more than simply make a virtual image visible. It must balance real-world transparency, display efficiency, color purity, and unwanted light leakage. The high-Q resonances of our nonlocal metasurface provide a way to address these requirements through precise control of both the spectrum and direction of light.

Beyond designing a device: designing the spectrum

The broader significance of this work extends beyond a single AR combiner. It demonstrates how two-dimensional topology optimization can serve as a powerful design framework for engineering nonlocal optical responses.

With the additional degrees of freedom provided by 2D optimization, we can move beyond simply shifting a resonance and instead determine which resonances should be retained, enhanced, or suppressed. This approach—working backward from a desired optical spectrum to discover the required nanostructure—could ultimately extend beyond AR to optical security, multispectral sensing, and other photonic technologies requiring precise spectral control.

Publication

C.-Y. Hsu, H.-T. Su, W.-T. Kuo, W.-Z. Li, Y.-T. Liu, Y.-C. Chang, and Y.-W. Huang*, "Two-dimensional topology optimized nonlocal metasurfaces for augmented reality," Nano Letters 26(12), 4080–4088 (2026) - LINK & PDF