Modern electromagnetic and optical/photonic devices often utilize electromagnetic or photonic band gap (EBG or PBG) structures to improve their performance. These structures usually consist of repetitions of standard unit cells with a small number of nonstandard cells. They can be designed using building blocks, similar to standard cells used in VLSI design. To improve design efficiency, modeling/design of EBG/PBG structures have been based on oversimplified lumped-element or transmission-line models that, strictly speaking, only valid for simple planar structures. For complicate or 3D structures, numerical simulators are always needed, which are however computationally expensive. For a structure with imperfection, a large domain is needed for accurate simulation, which is computationally prohibitive for any practical design.
In this project, a multi-block reduced order model (ROM) will be applied to develop an effective design methodology for EBG/PBG structures. The approach is able to reduce the computational time by several orders of magnitude with similar accuracy and resolution compared to numerical simulators. The methodology is an extension of a previous NSF-funded project for ROM thermal simulation of semiconductor chips. The proposed research will first select 2 sets of unit cells needed for each group of EBG and PBG structures. An efficient ROM for each select unit cell will be developed. To develop the ROM for each cell, a numerical simulator will be used to generate a large set of data accounting for parametric variations in each cell. These data are then used to calculate ROM parameters for the cell. This process involves extremely intensive computation and will be performed in an Nvidia GPU to improve the efficiency. Once the efficient ROM cells are generated, they can be stored in a library for constructing larger EBG/PBG structures for cost-effective design of electromagnetic/photonic devices.