Fast, Accurate Large-Scale Metasurface Optimization

Researchers in Stanford's Nanoscale and Quantum Photonics Lab have developed a fast and accurate method to analyze large area metasurfaces. Since traditional simulation techniques (e.g., finite difference time domain, finite element method) take too long and are prohibitively expensive, metasurface designers typically use approximations - optimizing each meta-atom in isolation from the others. To design metasurfaces with higher efficiencies, Stanford researchers developed software to quickly analyze the full metasurface. Using the transfer matrix method, a system of equations based on each meta-atom's input-output relationship is used to solve for electromagnetic fields capturing the interaction between meta-atoms (figure 1). The adjoint method optimizes the full metasurface. GPU nodes accelerate solving the matrix, and decrease simulation time and cost (figure 2). Quick gradient-based optimization iterations significantly improve overall device performance (figure 3). Stanford's quick, accurate and cost effective metasurface simulation facilitates the design of efficient metasurfaces for applications ranging from virtual and augmented reality devices to other imaging, sensors, and flat optical components.

Stage of Development –Prototype Software