Ultracompact Terahertz Spectrometer Based on Magneto‐Electric‐Coupled Optoelectronic Metasurface Assisted by Deep Learning

Miniaturized terahertz spectrometers hold revolutionary potential for portable real‐time spectral analysis, yet reports on such devices remain scarce. Existing compact spectrometers typically rely on microscale dispersive elements or tunable photodetector arrays, but these approaches face limitations in the terahertz regime due to their low photon energy and long wavelength. Here, a terahertz spectrometer framework is introduced by integrating dispersion and detection functionalities in one metasurface via magneto‐electric coupling processes. Such frameworks are self‐powered and in wavelength‐scale footprint, which apply to both point‐by‐point spectral reconstruction and computational one. Especially, assisted by deep learning techniques, spectral reconstructions with high spectral resolution and speed can be achieved, and over 15 000 spectra can be acquired in <1 min. By offering significance in compactness and acquisition efficiency, this strategy paves the way for miniaturized terahertz spectroscopy, novel functionalities of optoelectronics, and advanced integrated photonic platforms.