Zhongwei Jin

Breaking the diffraction limit in optical imaging is crucial for resolving subwavelength details in a wide range of applications, where super-oscillatory imaging and subtraction imaging are two common strategies for surpassing conventional resolution limits. We propose an end-to-end deep learning framework that integrates super-oscillatory focusing and subtraction imaging into a single jointly optimized vectorial Debye integral neural network pipeline, eliminating the traditional two-step acquisition and manual weighting process. With this end-to-end neural network, we further improve the focusing capability of the system to the sub-100-nm regime, enabling deep-subwavelength imaging resolution.

Bound states in the continuum (BICs) have garnered significant attention in recent years due to their ability to enhance light–matter interaction. The high quality ( Q ) factors of quasi-BICs make them particularly promising for applications such as lasing, sensing, and nonlinear harmonic enhancement. Here, we numerically and experimentally demonstrate dynamically tunable multiple quasi-BICs by integrating aluminum rod arrays with GST. By tailoring the geometric parameters of the aluminum rod arrays, multiple quasi-BICs are achieved, manifesting as distinct Fano resonances in the terahertz transmission spectrum. Upon coating the aluminum rod arrays with GST, both the resonant frequencies and linewidths of the quasi-BICs can be dynamically tuned through the phase transition of GST from the amorphous to the crystalline phase. Furthermore, these quasi-BICs exhibit strong electromagnetic field confinement, which can be significantly modulated via the GST phase transition. The dynamically tunable multiple quasi-BICs offer promising potential for applications in terahertz modulators, filters, and imaging.

Metasurfaces offer innovative approaches for manipulating electromagnetic waves at subwavelength scales. Recent advancements have focused on toroidal dipole (TD) and quasi-bound state in the continuum (quasi-BIC) modes, which are particularly attractive due to their capacity to enhance light-matter interaction. However, most metasurfaces with TD and quasi-BIC modes exhibit passive electromagnetic responses after fabrication, limiting their practical applications. This study presents both numerical and experimental investigations that demonstrate the active control of TD and quasi-BIC modes through the integration of symmetric and asymmetric aluminum dumbbell aperture arrays with the phase-change material Ge₂Sb₂Te₅ (GST). The symmetric hybrid dumbbell aperture array shows a pronounced TD response within the terahertz frequency range. In contrast, modifying the geometric parameters to disrupt the structural symmetry induces a quasi-BIC mode in the asymmetric hybrid dumbbell aperture array. Furthermore, as GST undergoes a phase transition from its amorphous to crystalline state, both TD and quasi-BIC modes become dynamically tunable, driven by changes in the conductivity of GST. Notably, significant modulation of the transmitted terahertz wave occurs around the frequencies corresponding to the TD and quasi-BIC modes during the GST phase transition. Symmetric and asymmetric hybrid dumbbell aperture arrays provide a versatile platform for generating tunable TD and quasi-BIC modes, with promising applications in terahertz modulators and filters.

Breaking the diffraction limit has been a key challenge in optical engineering and super-resolution imaging. In this work, we utilize a vectorial Debye integral neural network to design sub-diffraction focusing fields for high-NA objectives. By training the polarization states of incident light, we flexibly achieve transitions from diffraction-limited focusing to superoscillatory regimes. Through parameter adjustments, we optimize focal spot size, energy efficiency, and sidelobe distribution, achieving a focus with a 0.367λ FWHM and enhanced energy utilization. This method significantly simplifies the design process and demonstrates great potential for advanced optical applications, including super-resolution imaging and 3D field engineering.

Traditional superoscillation focusing (SOF) typically requires complex optimization of the incident light field. These complexities may limit the practical application of superoscillation. High-order radially polarized Laguerre-Gaussian beams inherently support SOF due to their multi-ring amplitude distribution and 0-π phase alternation, which align with the necessary destructive interference mechanisms. In this study, we demonstrate that by adjusting the beam’s mode order together with the incident beam size, we can easily control the full width at half maximum, field of view, and energy distribution of SOF. Moreover, high-order azimuthally polarized vortex-phase Laguerre-Gaussian beams can also achieve SOF, offering even better super-resolution effects. The distinct focusing behaviors of their circular components present unique opportunities for applications involving circular dichroism materials.

Metasurfaces consisting of subwavelength structures have shown unparalleled capability in light field manipulation. However, their functionalities are typically static after fabrication, limiting their practical applications. Though persistent efforts have led to dynamic wavefront control with various materials and mechanisms, most of them work in free space and require specialized materials or bulky configurations for external control. This deviates from the original intention of metasurface to realize compact and integrated devices. Here, we leverage the on-chip geometric metasurface associated with polarization reconfiguration of the guided wave, enabling three functions simultaneously: guided wave radiation, polarization multiplexing, and dynamic wavefront manipulation. We demonstrate proof-of-principle functionalities, including intensity-continuously tunable multifocal metalens, and dynamic zoom metalens as well as dynamic holography, based on a metasurface dressed lithium-niobate-on-insulator waveguide. Such an integrated platform for dynamic wavefront shaping implies the prospect of advancements in chip-integrated multifunctional meta-devices.

Recent advancements in nanotechnology have revolutionized terahertz (THz) technology. By enabling the creation of compact, efficient devices through nanoscale structures, such as nano-thick heterostructures, metasurfaces, and hybrid systems, these innovations offer unprecedented control over THz wave generation and modulation. This has led to substantial enhancements in THz spectroscopy, imaging, and especially bio-applications, providing higher resolution and sensitivity. This review comprehensively examines the latest advancements in nanoengineered THz technology, beginning with state-of-the-art THz generation methods based on heterostructures, metasurfaces, and hybrid systems, followed by THz modulation techniques, including both homogeneous and individual modulation. Subsequently, it explores bio-applications such as novel biosensing and biofunction techniques. Finally, it summarizes findings and reflects on future trends and challenges in the field. Each section focuses on the physical mechanisms, structural designs, and performances, aiming to provide a thorough understanding of the advancements and potential of this rapidly evolving technology domain. This review aims to provide insights into the creation of next-generation nanoscale THz devices and applications while establishing a comprehensive foundation for addressing key issues that limit the full implementation of these promising technologies in real-world scenarios.

The inter-plane crosstalk and limited axial resolution are two key points that hinder the performance of three-dimensional (3D) holograms. The state-of-the-art methods rely on increasing the orthogonality of the cross-sections of a 3D object at different depths to lower the impact of inter-plane crosstalk. Such strategy either produces unidirectional 3D hologram or induces speckle noise. Recently, learning-based methods provide a new way to solve this problem. However, most related works rely on convolution neural networks and the reconstructed 3D holograms have limited axial resolution and display quality. In this work, we propose a vision transformer (ViT) empowered physics-driven deep neural network which can realize the generation of omnidirectional 3D holograms. Owing to the global attention mechanism of ViT, our 3D CGH has small inter-plane crosstalk and high axial resolution. We believe our work not only promotes high-quality 3D holographic display, but also opens a new avenue for complex inverse design in photonics.

Resonant dielectric nanostructures have achieved significant advancements in the manipulation of light at the nanoscale. Particularly, bound states in the continuum (BICs) based on dielectric metasurfaces have greatly enhanced the intensity of light–matter interaction. However, most BICs in dielectric metasurfaces are fixed in their functionality once they are made. In this study, we present the development of switchable multiple quasi-BICs by combining dielectric nanostructures with vanadium dioxide. The resulting hybrid dielectric metasurface can support three types of BICs with different multipole origins for vanadium dioxide in the insulating phase. By introducing structural asymmetry through width adjustment, one quasi-BIC with a longitudinal toroidal dipole characteristic is excited under x-polarized incidence. Further, tuning the width allows for the generation of two additional quasi-BICs with distinct electromagnetic sources under y-polarized incidence. Additionally, the hybrid dielectric metasurface also supports a high-Q transverse toroidal dipole mode. Moreover, all quasi-BICs and toroidal dipole modes can be turned off when vanadium dioxide transitions into the metallic phase. The switchable multiple quasi-BICs hold promise for applications in optical modulators, tunable harmonic generation, and biosensors.