asasim: Adaptive sampling for electromagnetic simulations

For simulations of electromagnetic resonance spectra, where the locations of spectral features are unknown, and for wide-band simulations in general, a substantial number of wavelengths must be simulated for acceptable resolution, increasing computation time. This problem is exacerbated for spectra containing narrow-band features, as a high spectral resolution is required to even detect them.

To address this challenge, a heuristic algorithm is presented for electromagnetic simulations, which adaptively refines the local resolution of spectral features during a simulation. The method supports parallel processing and plugs in with existing simulation systems, such as (RCWA). It can routinely reduce the computational load by two orders of magnitude.

Program Summary: Program Title: asasim Program Files doi: http://dx.doi.org/10.17632/d6gty7kr2x.1 Licensing provisions: CC By 4.0 Programming language: MATLAB Nature of problem: Simulations are challenging when information is needed both on a long scale (broad interval) and on a short scale (high local resolution), such as wide-band electromagnetic spectra containing narrow-band features.

When resolution is insufficient, narrow-band features may be downright absent from the spectrum, if neighboring points are simulated on either side of a narrow peak. When local resolution is sufficient, it will necessarily be excessive in flat regions, wastefully increasing computation time. Solution method: The presented method enables adaptive resolution, which ensures that all peaks of a given minimum width are always detected and maximally resolved, while feature-less regions remain minimally resolved.

An optimum point spacing is derived for lorentzian peaks (descriptive of optical resonances) and is applied to optimize computation time.