The advent of metasurfaces in recent years has ushered in a revolutionary means to manipulate the behavior of light on the nanoscale. The design of such structures, to date, has relied on the expertise of an optical scientist to guide a progression of electromagnetic simulations that iteratively solve Maxwell's equations until a locally optimized solution can be attained. In this work, we identify a solution to circumvent this conventional design procedure by means of a deep learning architecture. When fed an input set of customer-defined optical spectra, the constructed generative network generates candidate patterns that match the on-demand spectra with high fidelity. This approach reveals an opportunity to expedite the discovery and design of metasurfaces for tailored optical responses in a systematic, inverse-design manner.

We study periodic inversion and phase transition of normal, displaced, and chirped finite energy Airy beams propagating in a parabolic potential. This propagation leads to an unusual oscillation: for half of the oscillation period the Airy beam accelerates in one transverse direction, with the main Airy beam lobe leading the train of pulses, whereas in the other half of the period it accelerates in the opposite direction, with the main lobe still leading - but now the whole beam is inverted. The inversion happens at a critical point, at which the beam profile changes from an Airy profile to a Gaussian one. Thus, there are two distinct phases in the propagation of an Airy beam in the parabolic potential - the normal Airy and the single-peak Gaussian phase. The length of the single-peak phase is determined by the size of the decay parameter: the smaller the decay, the smaller the length. A linear chirp introduces a transverse displacement of the beam at the phase transition point, but does not change the location of the point. A quadratic chirp moves the phase transition point, but does not affect the beam profile. The two-dimensional case is discussed briefly, being equivalent to a product of two one-dimensional cases.

controllability of light due to the limited flexibility rendered in periodic metastructures of simple unit cells. To overcome these deficiencies, metasurfaces comprised of multiple meta-atoms, such as gradient and multilayered metasurfaces, have been proposed and developed. [7][8][9] Relying on the collective effects of multiple meta-atoms, these metasurfaces present intriguing properties such as anomalous deflection, [7,10] arbitrary phase control, asymmetric polarization conversion, [8,11] wave-front shaping, [12][13][14] etc., which brings about extensive applications for imaging, optical signal processing, emission control, and much more. Here in our following discussion, we refer to unit cells composed of various meta-atoms as metamolecules, analogous to the hierarchical relationship between atoms and molecules in nature. In our definition of a metamolecule, we assume every two adjacent meta-atoms are not strongly coupled, in which case the overall properties of the metamolecule can be analytically predicted by the properties of its constituent meta-atoms. Such an assumption is valid in most metasurfaces that consist discrete, spatially variant building blocks.Despite the extraordinary properties of metasurfaces made up of metamolecules, designing multiple meta-atoms that collectively function as a device is a time-consuming task that requires labor-intensive trial-and-error simulations. The difficulty of the inverse design of such metamolecules arises from the intricate mechanisms of multistructured systems, the vast number of possible combinations of distinct meta-atoms, as well as the expensive 3D full wave simulations required. Traditionally, a practical solution to such a design follows three steps: 1) specifying a class of geometry with a few parameters as candidate meta-atoms, 2) carrying out parametric sweeps on these parameters, and 3) enumerating possible combinations of meta-atoms to meet the design objective. However, the limitation of the geometry in the strategy largely restricts the variety of the shapes of meta-atoms, which usually does not lead to an optimal solution, even after extensive and expensive simulations.Alongside the evolution of nanophotonics, various methods for expediting the design of photonic structures have been developed. Gradient-based adjoint methods, such as topology optimization, are a class of widely applied approaches for Molecules composed of atoms exhibit properties not inherent to their constituent atoms. Similarly, metamolecules consisting of multiple meta-atoms possess emerging features that the meta-atoms themselves do not possess. Metasurfaces composed of metamolecules with spatially variant building blocks, such as gradient metasurfaces, are drawing substantial attention due to their unconventional controllability of the amplitude, phase, and frequency of light. However, the intricate mechanisms and the large degrees of freedom of the multielement systems impede an effective strategy for the design and optimization of metamolecules. Here, a hybrid artificial-i...

Machine learning, as a study of algorithms that automate prediction and decision‐making based on complex data, has become one of the most effective tools in the study of artificial intelligence. In recent years, scientific communities have been gradually merging data‐driven approaches with research, enabling dramatic progress in revealing underlying mechanisms, predicting essential properties, and discovering unconventional phenomena. It is becoming an indispensable tool in the fields of, for instance, quantum physics, organic chemistry, and medical imaging. Very recently, machine learning has been adopted in the research of photonics and optics as an alternative approach to address the inverse design problem. In this report, the fast advances of machine‐learning‐enabled photonic design strategies in the past few years are summarized. In particular, deep learning methods, a subset of machine learning algorithms, dealing with intractable high degrees‐of‐freedom structure design are focused upon.

Designing complex physical systems, including photonic structures, is typically a tedious trial-and-error process that requires extensive simulations with iterative sweeps in multidimensional parameter space. To circumvent this conventional approach and substantially expedite the discovery and development of photonic nanostructures, here we develop a framework leveraging both a deep generative model and a modified evolution strategy to automate the inverse design of engineered nanophotonic materials. The capacity of the proposed methodology is tested through the application to a case study, where metasurfaces in either continuous or discrete topologies are generated in response to customer-defined spectra at the input. Through a variational autoencoder, all potential patterns of unit nanostructures are encoded into a continuous latent space. An evolution strategy is applied to vectors in the latent space to identify an optimized vector whose nanostructure pattern fulfills the design objective. The evaluation shows that over 95% accuracy can be achieved for all the unit patterns of the nanostructure tested. Our scheme requires no prior knowledge of the geometry of the nanostructure, and, in principle, allows joint optimization of the dimensional parameters. As such, our work represents an efficient, on-demand, and automated approach for the inverse design of photonic structures with subwavelength features.