An improved memory-efficient subdomain level discontinuous Galerkin time domain (SL-DGTD) method, which utilizes the repetition of the geometries, has been proposed to model the periodic and aperiodic structures with square uniform lattice. Significant reduction of memory consumption has been achieved comparing to the conventional SL-DGTD approach. Based on the previous pioneering work of memory-efficient SL-DGTD method, two improvements have been implemented to extend the scope of application. First, multiple kinds of cells, which can be placed arbitrarily on a square lattice, are allowed after a generalized interface recognition scheme. Therefore, besides the large finite periodic arrays successfully solved by previous memory-efficient SL-DGTD method, the aperiodic and/or the quasi-periodic array can be simulated with much less memory as well. Second, certain part of a cell can be cut out to form a new cell. In other words, the cells can be embedded hierarchically, and this is crucial for an efficient modeling of multiscale structures. The effectiveness and efficiency of these improvements has been validated via modeling both proof-of-idea cases and patch antenna arrays with satisfactory reduction in memory consumption.

中文翻译：

---

用于周期/准周期结构的改进的内存效率子域级不连续伽辽金时域方法

已经提出了一种改进的内存高效子域级不连续伽辽金时域 (SL-DGTD) 方法，该方法利用几何形状的重复，对具有方形均匀晶格的周期性和非周期性结构进行建模。与传统的 SL-DGTD 方法相比，显着减少了内存消耗。在之前内存高效SL-DGTD方法的开创性工作的基础上，进行了两项改进以扩展应用范围。首先，在通用的界面识别方案后，允许多种单元格，可以任意放置在方形点阵上。因此，除了先前内存效率高的 SL-DGTD 方法成功求解的大型有限周期阵列外，非周期和/或准周期阵列也可以用更少的内存进行模拟。其次，细胞的某些部分可以被切除以形成一个新的细胞。换句话说，细胞可以分层嵌入，这对于多尺度结构的有效建模至关重要。这些改进的有效性和效率已经通过对概念验证案例和贴片天线阵列进行建模并令人满意地减少了内存消耗得到了验证。