Fast Modeling of Finite Reflectarrays and Reconfigurable Intelligent Surfaces for Next-Generation Wireless Applications

Licentiate thesis, 2026

Reflectarrays and reconfigurable intelligent surfaces (RIS) enable low-profile, high-gain wavefront shaping with electrically large apertures in high-frequency wireless systems. However, accurate full-wave analysis of finite, conformal, and reconfigurable reflectarrays remains computationally expensive, especially when many configurations must be evaluated for design, tuning, optimization, or electromagnetic dataset generation. This thesis develops fast numerical modeling techniques for all-metal and cavity-backed reflectarrays that preserve local full-wave accuracy while accelerating array-level analysis. A hybrid Method-of-Moments–Physical-Optics (MoM–PO) approach is first introduced, where an all-metal reflectarray is represented by a perfect-electric-conductor (PEC) equivalent surface excited by MoM-derived aperture magnetic currents. A domain-decomposition framework is then developed for open-cavity-based array elements, in which unit cells are characterized separately and finite-array mutual coupling is modeled through over-the-air generalized admittance interactions. The formulation is further extended to conformal cavity-backed reflectarrays by coupling interior cavity and exterior radiation domains through generalized apertureadmittance operators. For electrically large and reconfigurable arrays, the thesis introduces a hybrid domain-decomposition/PO framework using characteristic-mode macro basis functions (MBFs). This enables efficient multi-state analysis by reusing pre-characterized element operators and reducing the number of aperture unknowns. Different MBF constructions, including characteristic mode analysis (CMA), the characteristic basis function method (CBFM), and singular value decomposition (SVD), are compared to quantify trade-offs between accuracy, compression, and conditioning. Finally, Woodbury-identity-based updates and Jacobi-preconditioned Generalized Minimal Residual (GMRES) solvers are used to accelerate recomputation under localized element changes. The resulting methods support scalable analysis, electromagnetic dataset generation, and data-driven optimization of reflectarray and RIS architectures.