Sub-diffraction terahertz backpropagation compressive imaging

Terahertz single-pixel imaging (TSPI) has garnered significant attention due to its simplicity and cost-effectiveness. However, the relatively long wavelength of THz waves limits sub-diffraction-scale imaging resolution. Although TSPI technique can achieve sub-wavelength resolution, it requires harsh experimental conditions and time-consuming processes. Here, we propose a sub-diffraction THz backpropagation compressive imaging technique. We illuminate the object with monochromatic continuous-wave THz radiation. The transmitted THz wave is modulated by prearranged patterns generated on the back surface of a 500-{\mu}m-thick silicon wafer, realized through photoexcited carriers using a 532-nm laser. The modulated THz wave is then recorded by a single-element detector. An untrained neural network is employed to iteratively reconstruct the object image with an ultralow compression ratio of 1.5625% under a physical model constraint, thus reducing the long sampling times. To further suppress the diffraction-field effects, embedded with the angular spectrum propagation (ASP) theory to model the diffraction of THz waves during propagation, the network retrieves near-field information from the object, enabling sub-diffraction imaging with a spatial resolution of ~{\lambda}0/7 ({\lambda}0 = 833.3 {\mu}m at 0.36 THz) and eliminating the need for ultrathin photomodulators. This approach provides an efficient solution for advancing THz microscopic imaging and addressing other inverse imaging challenges.