LWIR Imaging: AI-Designed Meta-Optics Overcome Challenges

Long-wavelength infrared (LWIR) imaging is crucial for applications ranging from night vision to defense. However, conventional lenses for these systems are bulky, expensive (often made from germanium), and limit device design.

Enter meta-optics: ultra-thin, lightweight alternatives based on nanostructured surfaces. While promising for size and weight reduction, meta-optics struggle with chromatic aberrations, where different wavelengths focus at different points. This has hindered their ability to replace traditional lenses, especially in the LWIR range.

Researchers have addressed this challenge with a new design framework called “MTF-engineering.” MTF (modulation transfer function) describes how well a lens preserves image detail across different spatial frequencies.

This framework tackles the limitations of broadband meta-optics, enabling the design and demonstration of LWIR meta-lenses in lab and real-world settings.  The team built upon existing techniques by optimizing the tiny nanopillars’ shape and arrangement on the meta-optic surface.

Key Innovations:

Deep Neural Networks (DNNs): Simulating light interaction with millions of nanopillars is computationally expensive. The team trained a DNN model to rapidly predict the phase shift caused by different pillar shapes, accelerating the design process.

Intuitive Figure of Merit (FoM): In typical inverse design, optimizing a complex FoM can lead to unclear results. Here, the FoM directly relates to the area under the MTF curve, maximizing the information that passes through the lens and creating high-quality images.

This combination of intuitive design principles and efficient computational methods significantly outperforms simple metalenses.

The fabricated meta-optics were made from silicon, opening the door to germanium-free LWIR systems. While further improvements are needed to match commercial lenses, this research represents a significant step forward.

The team has even publicly made their “meta box” framework available, inviting the broader scientific community to leverage it for their meta-optics development. This paves the way for a future filled with innovative, high-performance LWIR imaging devices enabled by meta-optics.

Read more

Related Content: IR Detector – Dual-Band, CMOS-Compatible, Bias-Tunable