ENTERPRISE AI ANALYSIS

A golay metalens for long-range, large aperture, thermal imaging via sparse aperture computational imaging

Mid-wave infrared (MWIR) imaging is known for its sensitivity to temperature variations and ability to penetrate atmospheric haze, which makes it effective for target detection and tracking, remote 3D imaging, and astronomical observations. Most of these applications, however, necessitate long-range imaging with large aperture optics, and traditional MWIR refractive lenses become increasingly heavy and bulky as the aperture size grows. Metalenses are ultrathin optical elements that focus light through subwavelength structures, offering lightweight designs, potential freedom from spherical aberrations, and complex phase customization, providing more degrees of freedom than traditional lenses, enabling various applications. Initial meta-optics were fabricated primarily using electron beam lithography with the size limited to ~10^3-10^4 λ^2. Advances in nano-fabrication have enabled apertures of ~10^5 λ^2 through deep ultraviolet lithography, nanoimprinting, and direct laser writing. The largest reported metalens (100 mm diameter), for example, was created by stitching multiple exposure fields with different photolithography reticles during each exposure cycle, leading to aberrations and efficiency decrease compared to single-shot photolithography. Scaling such large metalenses to even larger apertures remains extremely challenging. A promising approach for building large aperture systems involves using sparse aperture configurations, exemplified by the Atacama Large Millimeter Array, a 66-telescope array where smaller units combine to achieve the resolution of a larger aperture. In optical sparse aperture systems, electromagnetic fields directly interfere on a single focal plane detector. Applying metalenses in a sparse aperture configuration helps overcome size limitations while maintaining a lightweight design. However, design and fabrication challenges remain due to their differences from traditional refractive lenses, leading current research to still be in its early stages. This includes simulation studies and experimental demonstrations of simple configurations with up to 4 sub-apertures, all at the micron scale, which are unsuitable for long-range imaging. For such applications, larger apertures with more sub-apertures in complex arrangements, ranging from centimeter to meter scale, are recommended. Sparse apertures generate large, asymmetric point spread functions (PSFs) leading to perceptually blurry images, and metalenses have lower efficiencies, causing reduced signal-to-noise ratio (SNR) than standard lenses. Thus, computational reconstruction is essential for these sparse aperture metalens systems, whose current works employ only basic techniques like Wiener filtering and the Richardson-Lucy algorithm, resulting in suboptimal outcomes. State-of-the-art deep learning-based image reconstruction methods, effective in addressing image degradation and noise have yet to be implemented. Here, we proposed a computational imaging system referred to as Golay metalens. Modified from the Golay array configuration, which targets multiple sub-apertures and uses compact non-redundant autocorrelations to maximize spatial frequency bandwidth, our system features an 89 mm outer diameter with seven sub-apertures, operating in the MWIR (see Fig. 1a for system schematic). While the weight of the plano-convex lens scales as the cube of the focal length, the metalens scales only as a square, leading to an increasing weight disparity as focal length grows (Fig. 1b). Our Golay metalens realized a 15-fold weight reduction compared to an equivalent traditional refractive lens. Fabricated on an all-silicon platform using cost-effective direct laser writing, our system utilizes an effective image reconstruction method based on the half-quadratic splitting (HQS) technique with a deep learning-based denoiser prior, acquiring high-resolution images for objects with MWIR radiation. To address the need for larger apertures for higher spatial resolution at longer distances (Fig. 1c), we also designed and simulated a large recursive Golay metalens (outer diameter of 495.1 mm), formed by arranging 7 Golay6 +1 metalenses as sub-apertures in a Golay6 + 1 configuration. Our Golay6 +1 and Recursive Golay6 + 1 metalenses overcome the fabrication constraint on the maximum diameter of a single metalens (d), achieving a larger outer diameter (D) and offering scalable resolution enhancement factors (D/d) of 5.56 and 30.94, respectively. Taken together, our study paves the way for developing lightweight, cost-effective, and high-resolution imaging systems suitable for long-range MWIR applications.