Hyperspectral terahertz microscopy via nonlinear ghost imaging

Ghost imaging, based on single-pixel detection and multiple pattern illumination, is a crucial investigative tool in difficult-to-access wavelength regions. In the terahertz domain, where high-resolution imagers are mostly unavailable, ghost imaging is an optimal approach to embed the temporal dimension, creating a “hyperspectral” imager. In this framework, high resolution is mostly out of reach. Hence, it is particularly critical to developing practical approaches for microscopy. Here we experimentally demonstrate time-resolved nonlinear ghost imaging, a technique based on near-field, optical-to-terahertz nonlinear conversion and detection of illumination patterns. We show how space–time coupling affects near-field time-domain imaging, and we develop a complete methodology that overcomes fundamental systematic reconstruction issues. Our theoretical-experimental platform enables high-fidelity subwavelength imaging and carries relaxed constraints on the nonlinear generation crystal thickness. Our work establishes a rigorous framework to reconstruct hyperspectral images of complex samples inaccessible through standard fixed-time methods.