Objective-driven metasurface design
Objective-driven metasurface design
Embedded Files
A deep learning modeling approach is introduced that significantly improves both speed and accuracy compared to conventional techniques used to characterize the subwavelength optical structures. It overcomes two key challenges that have limited previous neural-network-based design schemes: input/output vector dimensional mismatch, and accurate EM-wave phase prediction. Combined with optimization algorithms or neural networks, this approach can be generically applied to a wide variety of metasurface device designs
Deep learning modeling approach for metasurface with high degrees of freedom
Deep learning modeling approach for metasurface with high degrees of freedom
We demonstrate a deep learning-based metasurface modeling approach for accurate spectrum response prediction. This approach can handle a significantly larger set of input parameters, including various shapes, thickness, lattice size and refractive index of the meta-atoms. This network can be adopted as an efficient modeling tool in various application scenarios, including rapid meta-atom design and meta-device optimization.
Target Design by GAN
Generative network for freeform metasurface design
Generative network for freeform metasurface design
We have proposed a metasurface design network based on the GAN (Generative Adversarial Networks) architecture that is capable of efficiently producing numerous multifunctional meta-device designs on demand. This work demonstrates several important milestones as 1) the first free-form all-dielectric meta-atom design network; 2) the first free-form multifunctional metasurface design network and 3) the first metasurface lens designed by GANs.
Simulated E-field
Metasurface design Local magnification
AI to predict mutual coupling effects in metasurfaces
AI to predict mutual coupling effects in metasurfaces
In traditional meta-atom design approaches, unit cell boundary conditions were adopted during full-wave simulations. However, in real metasurface designs, each meta-atom is usually surrounded by non-identical meta-atoms, for which near-field coupling effects will differ from those used to calculate the original response. To address this issue, we have proposed a deep learning network that accounts for mutual coupling effects to predict the local responses of target meta-atoms. The fully-trained network takes the dimension of a target meta-atom and its neighbors as input and generates its accurate local response in milliseconds.
DNN-enabled active metasurface embedded design
DNN-enabled active metasurface embedded design
Design of active metasurface structures can be a computation-intensive task given the massive number of degrees of freedom encompassing free-form meta-atom geometries. We have developed a design approach for complex photonic structures involving embedded phase-change material metasurfaces inside a multilayer cavity and have applied this methodology to generate tunable mid-wave infrared bandpass filters.
Page updated
Google Sites
Report abuse