Large-aperture high-performance achromatic thermal imaging with a dispersive metalens empowered by a spectrum-informed computational framework

【摘要】 The growing demand for compact, lightweight, and cost-effective long-wave infrared(LWIR) optical systems has driven interest in metalenses. However, their performance is limited by the aperture-dispersion trade-off. Existing dispersion compensation strategies are limited by the constrained phase modulation ranges and dispersions of these unit cells, making it difficult to simultaneously achieve large apertures, broad operational bandwidths, and high numerical apertures. To address this, we propose a spectrum-informed computational deblurring framework for correcting chromatic aberrations in LWIR metalens imaging. The first stage employs a Wiener filtering network for preliminary correction, followed by a U-Net architecture for refined image restoration. We demonstrate this approach using a 2-cm aperture, 0.45-NA(numerical aperture) ultrathin metalens(0.4-mm thickness, 0.38-g mass) fabricated via standard deep silicon etching techniques. Experimental validation reveals significant improvements in image quality compared to the unprocessed image: the peak signal-to-noise ratio(PSNR)increases from 13.73 to 21.90 dB, and the structural similarity index measure(SSIM) improves from 0.496 to 0.836. These results effectively overcome limitations of conventional dispersion compensation methods, resolving the aperture-dispersion tradeoff. The proposed ultrathin imaging system offers a transformative pathway for next-generation portable LWIR devices, with strong potential in applications such as security surveillance, industrial inspection, and mobile thermal sensing.更多还原

【Abstract】 The growing demand for compact, lightweight, and cost-effective long-wave infrared(LWIR) optical systems has driven interest in metalenses. However, their performance is limited by the aperture-dispersion trade-off. Existing dispersion compensation strategies are limited by the constrained phase modulation ranges and dispersions of these unit cells, making it difficult to simultaneously achieve large apertures, broad operational bandwidths, and high numerical apertures. To address this, we propose a spectrum-informed computational deblurring framework for correcting chromatic aberrations in LWIR metalens imaging. The first stage employs a Wiener filtering network for preliminary correction, followed by a U-Net architecture for refined image restoration. We demonstrate this approach using a 2-cm aperture, 0.45-NA(numerical aperture) ultrathin metalens(0.4-mm thickness, 0.38-g mass) fabricated via standard deep silicon etching techniques. Experimental validation reveals significant improvements in image quality compared to the unprocessed image: the peak signal-to-noise ratio(PSNR)increases from 13.73 to 21.90 dB, and the structural similarity index measure(SSIM) improves from 0.496 to 0.836. These results effectively overcome limitations of conventional dispersion compensation methods, resolving the aperture-dispersion tradeoff. The proposed ultrathin imaging system offers a transformative pathway for next-generation portable LWIR devices, with strong potential in applications such as security surveillance, industrial inspection, and mobile thermal sensing.更多还原