Alegret Joan, Käll Mikael, Johansson Peter
Department of Applied Physics, Chalmers University of Technology, S-412 96 Göteborg, Sweden.
Phys Rev E Stat Nonlin Soft Matter Phys. 2007 Apr;75(4 Pt 2):046702. doi: 10.1103/PhysRevE.75.046702. Epub 2007 Apr 6.
We present a computational algorithm which speeds up Green's tensor nano-optics calculations by means of optimizing the mesh that represents the system we want to investigate. The algorithm automates the process of creating a variable-size mesh that describes an arbitrary nanostructure. The total number of elements of this mesh is smaller than that of a regular mesh representing the same structure, and thus the Green's tensor calculations can be performed faster. Precision, however, is kept at a similar level than for the regular mesh. Typically, the algorithm yields a mesh that speeds up Green's tensor calculations by a factor of 4, while giving a maximum error in the field magnitude of about 5%. The speed-up factor makes it very suitable for otherwise lengthy calculations, and the error should be acceptable for most applications.
我们提出了一种计算算法，该算法通过优化表示我们想要研究的系统的网格来加速格林张量纳米光学计算。该算法自动创建一个可变大小的网格，该网格描述任意纳米结构。此网格的元素总数比表示相同结构的规则网格的元素总数少，因此可以更快地执行格林张量计算。然而，精度与规则网格保持在相似水平。通常，该算法生成的网格可将格林张量计算速度提高4倍，同时在场强方面给出约5%的最大误差。加速因子使其非常适合于原本冗长的计算，并且该误差对于大多数应用来说应该是可以接受的。
