Neural network-based surrogate model for inverse design of metasurfaces

Metasurfaces composed of spatially arranged ultrathin subwavelength elements are promising photonic devices for manipulating optical wavefronts,with potential applications in holography,metalens,and multiplexing communications.Finding microstructures that meet light modulation requirements is always a challenge in designing metasurfaces,where parameter sweep,gradient-based inverse design,and topology optimization are the most commonly used design methods in which the massive electromagnetic iterations require the design computational cost and are sometimes prohibitive.Herein,we propose a fast inverse design method that combines a physicsbased neural network surrogate model(NNSM)with an optimization algorithm.The NNSM,which can generate an accurate electromagnetic response from the geometric topologies of the meta-atoms,is constructed for electromagnetic iterations,and the optimization algorithm is used to search for the on-demand meta-atoms from the phase library established by the NNSM to realize an inverse design.This method addresses two important problems in metasurface design:fast and accurate electromagnetic wave phase prediction and inverse design through a single phase-shift value.As a proof-of-concept,we designed an orbital angular momentum(de)multiplexer based on a phase-type metasurface,and 200 Gbit/s quadrature-phase shift-keying signals were successfully transmitted with a bit error rate approaching 1.67×10^(-6).Because the design is mainly based on an optimization algorithm,it can address the“one-to-many”inverse problem in other micro/nano devices such as integrated photonic circuits,waveguides,and nano-antennas.

Orbital angular momentum mode logical operation using optical diffractive neural network

Optical logical operations demonstrate the key role of optical digital computing,which can perform general-purpose calculations and possess fast processing speed,low crosstalk,and high throughput.The logic states usually refer to linear momentums that are distinguished by intensity distributions,which blur the discrimination boundary and limit its sustainable applications.Here,we introduce orbital angular momentum(OAM)mode logical operations performed by optical diffractive neural networks(ODNNs).Using the OAM mode as a logic state not only can improve the parallel processing ability but also enhance the logic distinction and robustness of logical gates owing to the mode infinity and orthogonality.ODNN combining scalar diffraction theory and deep learning technology is designed to independently manipulate the mode and spatial position of multiple OAM modes,which allows for complex multilight modulation functions to respond to logic inputs.We show that few-layer ODNNs successfully implement the logical operations of AND,OR,NOT,NAND,and NOR in simulations.The logic units of XNOR and XOR are obtained by cascading the basic logical gates of AND,OR,and NOT,which can further constitute logical half-adder gates.Our demonstrations may provide a new avenue for optical logical operations and are expected to promote the practical application of optical digital computing.