Inverse design in nanophotonics

Abstract

Recent advancements in computational inverse design have begun to reshape the landscape of structures and techniques available to nanophotonics. Here, we outline a cross section of key developments at the intersection of these two fields: moving from a recap of foundational results to motivation of emerging applications in nonlinear, topological, near-field and on-chip optics.The development of devices in nanophotonics has historically relied on intuition-based approaches, the impetus for which develops from knowledge of some a priori known physical effect. The specific features of such devices are then typically calculated and matched to suitable applications by tuning small sets of characteristic parameters. This approach has had a long track record of success, giving rise to a rich and widely exploited library of templates that includes multilayer thin films 1 , Fabry-Perot 2 and microring resonators 3 , silicon waveguides 4,5 , photonic crystals 6 , plasmonic nanostructures 7 , and nanobeam cavities 8 , top of Fig. 1. Combining the principles of index guiding and bandgap engineering, along with material resonances, this collection of designs enables remarkable manipulation of light over bands of frequencies spanning from the ultra-violet to the mid infrared: group velocity can be slowed by more than two orders of magnitude 9 , light confined to volumes thousands of times smaller than its free-space wavelength 10 , and resonances made to persist in micron sized areas for tens of millions of cycles 11 .Yet, as the scope of nanophotonics broadens to include large bandwidth or multi-frequency applications, nonlinear phenomena, and dense integration, continuing with this prototypical approach poses a challenge of increasing complexity. For instance, consider the design of a wavelength-scale structure for enhancing nonlinear interactions 12 , discussed below. Even in the simplest case, several interdependent characteristics must be simultaneously optimized, among which are large quality factors at each individual wavelength and nonlinear overlaps, which must be controlled in as small a volume as possible. In such a situation, the templates of the aforementioned standard library offer no clear or best way to proceed; there is no definite reason to expect that an optimal design can be found in any of the traditional templates, or that such a design necessarily exists. Moreover, the performance of a given nonlinear device is likely to be highly dependent on the particular characteristics of the problem, and as greater demands are placed on functionality it becomes increasingly doubtful that any one class of structures will have the broad applicability of past devices. This lack of evident strategies for broadband applications also brings to attention the space of structures included in the standard photonic library. Predominately, traditional designs are repetitive mixtures and combinations of highly symmetric shapes described by a small collection of parameters. Since intuition-based optimization is then carried out ...

Compounding Meta‐Atoms into Metamolecules with Hybrid Artificial Intelligence Techniques

“…However, photonic media with identical meta-atoms may offer only inadequate optimizing and designing nanophotonic devices. [15][16][17][18][19][20] To identify the optimized parameters of a device, the algorithm computes the gradient, or sensitivity, through the corresponding adjoint problem and updates the parameters along the deepestgradient direction. In addition to adjoint methods, genetic algorithms and related variations also play important roles in the design of photonic structures.…”

A Bidirectional Deep Neural Network for Accurate Silicon Color Design

“…A deep learning approach can analyze the scattering spectra of silicon nanostructure within the diffraction‐limited area and output digital information, which breaks the limits of optical information storage and already beats the Blu‐ray Disk technology . In the field of nanophotonics, computational inverse design can reshape the landscape and techniques available to complex and emerging applications . Recent advancements in deep neural networks (DNNs) have demonstrated efficient forward‐modeling that can predict resonance spectrum accurately, and perform the inverse design of photonic device structures .…”

“…[8] In the field of nanophotonics, computational inverse design can reshape the landscape and techniques available to complex and emerging applications. [9] Recent advancements in deep neural networks (DNNs) have demonstrated efficient forward-modeling that can predict resonance spectrum accurately, and perform the inverse design of photonic device structures. [10][11][12][13][14][15][16][17][18] The general steps usually involve a one-time investment of sufficient EM simulation data, which are composed of variable device parameters and corresponding optical resonance at different wavelengths, followed by constructing DNNs.…”

Inverse Design of Photonic Topological Insulators with Extra‐Wide Bandgaps

“…Topology optimization was originally proposed in elastic fields and recently used to design novel PCs with exotic properties . However, topology optimization for the design of topological PCs is still challenging.…”

“…Topology optimization was originally proposed in elastic fields [39,40] and recently used to design novel PCs with exotic properties. [41][42][43][44][45][46][47][48][49][50][51][52][53][54] However, topology optimization for the design of topological PCs is still challenging. Different from exploiting the band folding mechanism in the conventional approach, [19][20][21]36] we directly focus on forming double Dirac cones, degenerated by dipolar (D) and quadrupolar (Q) modes, at the Γ point at any desired frequency by topology optimization.…”