Tunable Nonlocal Metasurfaces for Edge Computing and Processing

The contract seeks the development of tunable nonlocal metasurfaces designed to perform front-end analog image processing directly within the optical path of mid-wave infrared thermal-imaging systems, aiming to replace or reduce reliance on digital post-processing. Current thermal imaging systems suffer from high latency, power consumption, and complexity due to their dependence on computational methods for tasks like edge detection and spatial filtering. The proposed devices must leverage patterned metasurfaces fabricated from practical MWIR-compatible materials such as silicon or gallium arsenide, enabling scalable manufacturing while operating under realistic partially coherent illumination conditions—unlike prior lab-based demonstrations limited to coherent sources. Successful designs must account for the effects of source coherence on key performance metrics including contrast, throughput, signal-to-noise ratio, and numerical aperture compatibility, establishing predictive design rules that link optical parameters to system behavior. Preference will be given to solutions offering broadband MWIR operation, minimal size weight and power burden, and clear pathways for integration into field-deployable sensing platforms, with special emphasis on electro-optical tunability for real-time adaptive processing. The solicitation, titled Tunable Nonlocal Metasurfaces for Edge Computing and Processing under OSW26BZ04-DV011, is a small business set-aside targeting entities with fewer than 500 employees and is administered by the Office of the Secretary of Defense within the Department of Defense. Proposals are due by July 22, 2026, and must demonstrate proof-of-concept devices operating under realistic thermal imaging conditions, not just controlled laboratory environments. The technology must bridge the gap between theoretical metasurface designs and practical, deployable systems, ensuring compatibility with existing thermal imager architectures. While fabrication methods should enable mass production, the core innovation lies in the ability to perform complex optical computations—analogous to convolution or differentiation—directly in the optical domain without digitization, significantly lowering computational load downstream. The ultimate goal is to enable faster, more power-efficient, and more compact thermal imaging systems with real-time adaptive capabilities through dynamically tunable metasurface elements.