Adjoint Optimization of 3D Printed RF GRIN Lenses

Abstract

Additive manufacturing makes possible the rapid design and testing of gradient index (GRIN) lenses. GRIN lenses, whose refractive index profiles are spatially-varying, have been shown to improve the gain of antenna systems while in more compact forms. However, optimization of the GRIN lens profiles is difficult because the lenses can be many wavelengths wide, and the simulations slow to compute. Furthermore, the large number of design variables makes solving the GRIN lens profile a challenging problem. However, these problems can be remedied using adjoint sensitivities to compute the gradient for every design variable in only two simulations. Then local optimization approaches can follow these gradients to rapidly converge to a good candidate solution. To demonstrate the efficacy of the approach, a GRIN lens design was optimized to have the same measured broadside gain as a 20 dB standard gain horn.