AI enhanced strong-field terahertz spectral detection and imaging

Terahertz (THz) waves demonstrate advantages including broad absolute bandwidth, low photon energy, and superior penetration capability, showing significant potential in nondestructive testing and spectroscopic imaging applications. However, current techniques face challenges including unknown material refractive indices, pulse signal aliasing, and time-consuming detection processes, which hinder the advancement and commercialization of THz-based nondestructive testing. This work presents a nondestructive testing system employing an enhanced 4-inch spintronic strong-field THz emitter, integrated with neural-network-assisted thickness prediction and contour detection. The system achieves micrometer-scale accuracy for ultrathin materials and rapid defect identification in large-scale samples. Experimental results demonstrate 99.8% measurement accuracy within ±8 μm for micrometer-scale samples, with submillimeter-level depth resolution for defect characterization and contour imaging. This technology demonstrates scalability and high flexibility, laying the foundation for thickness measurement of multilayer composite coatings and internal defect detection in complex materials, showing broad prospects in surface coating processing, component maintenance, and cultural heritage preservation.