Compelling evidence suggests that the interaction between electromagnetic metasurfaces and deep learning gives rise to the proliferation of intelligent metasurfaces in the past decade. In general, deep learning offers a transformative force to reform the design and working style of metasurfaces. A majority of the inverse-design literature announce that, given a user-defined input, the pre-trained deep learning models can quickly output the metasurface candidates with high fidelity. However, they largely ignore an important fact, that is, the practical input is always semi-known. In this work, we introduce a generation-elimination network that is robust to semi-known input and information pollution. The network is composed of a generative network to generate a number of possible answers and then a discriminative network to eliminate suboptimal answers. We benchmark the feasibility via two scenes, the on-demand metasurface design of the reflection spectra and the far-field pattern. In the microwave experiment, we fabricated and measured the reconfigurable metasurfaces to automatically meet the semi-known beam steering requirement that widely exist in wireless communication. Our work for the first time answers the question of how to cope with semi-known input, which is ubiquitous in a panoply of real-world applications, such as imaging, sensing, and communication across noisy environment.

In recent years, topological states in photonic artificial structures have attracted great attention due to their robustness against certain disorders and perturbations. To readily understand the underlying principles, topological edge modes in one-dimensional (1D) system have been widely investigated, which bring aboutthe discovery of novel optical phenomena and devices. In this article, we review our recent advances in topological edge modes. We introduce the connection between topological orders and effective electromagnetic parameters of photonic artificial structures in band gaps, discuss experimental demonstration of robust topological modes and their potential applications in wireless power transfer, sensing and field of optics, and give a brief introduction of future opportunities in 1D topological photonics.

As one of the most important technologies for the next generation very-large scale integrated circuit fabrication, extreme ultraviolet (EUV) lithography has attracted more and more attention in recent years. However, in EUV lithography, the optical distortion of the printed image on wafer always has negative impacts on the imaging performance. Thus, to enhance the imaging performance of EUV system, especially for small critical dimensions, in this work, a novel optical proximity correction (OPC) system based on the deep learning technique is proposed. It includes a forward module and an inverse module, where the forward module is employed to fast and accurately map the mask to the corresponding near field of the plane above the stack to help the construction of training dataset for the inverse module operation, and the inverse module is employed to fast and accurately map the target printed image to the corrected mask. Numerical examples demonstrate that compared with traditional full-wave simulation, the forward module can greatly improve the computational efficiency including the required running time and memory. Meanwhile, different from time consuming iterative OPC methods, the corrected mask can be immediately obtained as the target printed image is input using the trained inverse module.

Photonic integrated circuits (PICs) and microwave integrated circuits (MICs) have been widely studied, but both of them face the challenge of miniaturization. On one hand, the construction of photonic elements requires spaces proportional to wavelength, and on the other hand, electromagnetic compatibility issues make it challenging to reach high-density layouts for MICs. In this paper, we review the research advances of miniaturized PICs and MICs based on surface plasmon polaritons (SPPs). By introducing SPPs, miniaturized photonic elements at subwavelength scales are realized on PICs, which can be used for highly integrated interconnects, biosensors, and visible light wireless communications. For MICs, since the metals behave as perfect conductors rather than plasmonic materials at microwave frequencies, plasmonic metamaterials are proposed to support spoof SPPs. Spoof SPPs possess similar characteristics to SPPs and can be used to realize high-density channels on MICs. Moreover, combining the latest theoretical research on SPPs, future tendencies of SPP-based MICs are discussed as well, including further miniaturization, digitization, and systematization.

Bessel beam is an important propagation-invariant optical field. The size and shape of its central spot remain unchanged in the long-distance transmission process, which has a wide application prospect. In this paper, we find that zero-index media (ZIM) metalen can be designed to realize the unique Bessel beam. On the one hand, based on the metal-dielectric multilayered structure with sub-wavelength unit cells, the anisotropic epsilon-near-zero media (ENZ) metalen is proposed for generating the robust Bessel beam, which is immune to the defects placed in the transmission path or the inside of the structure. The ZIM metalens uncover that ENZ media provide a new way to generate Bessel beams beyond the conventional convex prisms. On the other hand, with the help of the uniform field distribution of ZIM, enhanced (multi-channel) Bessel beams based on multiple point sources (exit surfaces) are studied in the isotropic ENZ metalens. In addition, the Bessel beam generated by the ZIM metalen has also been extend to the epsilon-mu-near zero metamaterial realized by two dimensional photonic crystals. Our results not only provide a new way to generate Bessel beam based on the ZIM metalens, but also may enable their use in some optical applications, such as in fluorescence microscopy imaging, particle trapping, and wave-front tailoring.

Metamaterials offer great promise for engineering electromagnetic properties beyond the limits of natural materials. A typical example is the so-called spoof surface plasmons (SPs), which mimic features of optical SPs without penetrating metal at lower frequencies. Spoof SPs inherit most of the properties of natural SPs, including dispersion characteristics, field confinement, localized resonance, and subwavelength resolution, and therefore are highly expected to offer a new solution for low-frequency applications. With the development of spoof SPs, three different theories have been introduced. The first one is the description of subwavelength corrugated metal surfaces by a metamaterial that hosts an effective plasma frequency. The second one is developed with high-index contrast grating, which can realize propagation with ultra low loss and localization with ultrahigh Q-factor resonance. The last one is structural dispersion induced SPs, a perfect low-frequency analogue of optical SPs, realized by exploiting the well-known structural dispersion waveguide modes only with positive-ɛ materials. Here, the developments of these three theories including propagation and localized SPs are reviewed, focusing primarily on the fundamental and representative applications.

In this work, we investigate the feasibility of applying deep learning to phase synthesis of reflectarray antenna. A deep convolutional neural network (ConvNet) based on the architecture of AlexNet is built to predict the continuous phase distribution on reflectarray elements given the beam pattern. The proposed ConvNet is sufficiently trained with data set generated by array-theory method. With radiation pattern and beam direction arrays as input, the ConvNet can make real-time and fairly accurate predictions in milliseconds with the average relative error below 0.7%. This paper shows that deep convolutional neural networks can ``learn'' the principle of reflectarray phase synthesis due to their inherent powerful learning capacity. The proposed approach may provide us a potential scheme for real-time phase synthesis of antenna arrays in electromagnetic engineering.

Exceptional points are spectral singularities in non-Hermitian systems at which two or more eigenvalues and their corresponding eigenvectors simultaneously coalesce. Originating from quantum theory, exceptional points have attracted significant attention in optics and photonics because their emergence in systems with nonconservative gain and loss elements can give rise to many counterintuitive phenomena. Metasurfaces - two-dimensional artificial electromagnetic materials structured at the subwavelength scale - can provide a versatile platform for exploring such non-Hermitian phenomena through the addition of dissipation and amplification within their unit cells. These concepts enable a wide range of exotic phenomena, including unidirectional propagation, adiabatic mode conversion, and ultrasensitive measurements, which can be harnessed for technological applications. In this article, we review the recent advances in exceptional-point and non-Hermitian metasurfaces. We introduce the basic theory of exceptional point and non-Hermiticity in metasurfaces, highlight important achievements and applications, and discuss the future opportunities of non-Hermitian metasurfaces from basic science to emerging technologies.

A low loss propagating electromagnetic wave is shown to exist at a gradual interface between two lossy conductive media. Such a surface wave may be guided by a seafloor-seawater interface and it may be used in radio communication and imaging underwater. It should allow communication distances of the order of 500 m at 10 kHz along a sandy seabed. Similar surface waves may also be guided by various tissue boundaries inside a human body. For example, such surface wave solutions may exist at planar interfaces between skull bones and grey matter inside a human head at 6 GHz.