Abstract Photonic computing has emerged as a promising platform for accelerating computational tasks requiring high degrees of parallelism, ranging from signal and image processing to neural network tasks. This study presents meta‐DFT (discrete Fourier transform), a single‐layer metasurface device, designed to perform optical complex‐to‐complex DFT with digital time complexity. A key challenge in free‐space optical computing lies in digital error control. This approach addresses this by spatially separating light into discrete focal spots, enabling complex phase reconstruction via an interferometric method with an integrated reference metalens, alongside an error mitigation scheme. The device's performance is systematically evaluated using input vectors with random complex amplitudes and phases, demonstrating error mitigation capability to reduce the error by half with as few as five pixels per output focal spot. These findings pave the way toward the advancement of metasurface‐based optical computation, offering a robust framework readily extensible to arbitrary complex‐valued matrix‐vector multiplication (MVM).
Tanuwijaya et al. (Mon,) studied this question.