The Yee algorithm for electromagnetic simulations is widely known to have many advantages, including the following crucial two: (i) Its calculations are local and therefore efficiently parallelizable--enabling simulations that capitalize on the speed and scalability of high-performance computing architecture. (ii) Yee's method faithfully preserves the symplectic geometry of Maxwell's equations, improving its accuracy in long-time numerical simulations. Whereas previous geometric generalizations of Yee's method have sacrificed its scalability, in this article the Yee algorithm is generalized to higher order and unstructured meshes in a manner that fully preserves both its scalability and geometric naturalness. This generalization is achieved by prioritizing the locality of the algorithm, reflecting the physical locality of Maxwell's equations. Specifically, we demonstrate that Yee's method is but a special case of a larger family of symplectic, finite element exterior calculus (FEEC) methods that use scalable, local approximations of mass matrices. We discuss the numerical advantages of this family of methods, which we call scalable FEEC (SFEEC) methods.
翻译:众所周知,电磁模拟的Yee算法有许多优点,包括以下两个关键优点:(一)它的计算是局部的,因此是可有效平行的、可有效平行的模拟,利用高性能计算结构的速度和可伸缩性。 (二)Yee的方法忠实地保留了Maxwell方程式的截面几何学,提高了其长期数字模拟的准确性。虽然以前Yee方法的几何概括性已经牺牲了其可伸缩性,但在本条中,Yee算法被普遍地推广到更高的顺序和不结构的meshes,从而充分保存其可伸缩性和几何性自然性。通过确定算法地点的优先顺序,反映Maxwell方程的物理位置,实现了这一普遍化。具体地说,我们证明Yee的方法只是使用可伸缩缩放的、本地矩阵近似值的方法的较大组合的一个特殊例子。我们讨论了这一组方法的数值优势,我们称之为伸缩式法。