Constrained Genetic Disentangling of Close Binary Spectra

Abstract

The spectral disentangling technique makes possible separation of individual component spectra in binary or multiple systems, and determination of the orbital elements in a self-consistent way. Since its introduction, a number of variants of their basic idea have been implemented. We present yet another ‘direct’ approach using optimization by genetic algorithm. Starting with an initial random flux distribution representing individual spectra, genetic optimization returns both individual component spectra and an optimal set of orbital parameters only constrained by time-series of the observed composite spectra of the binary system. Benchmark tests on V453 Cyg, which is an eclipsing binary with total eclipse, as well as tests on the artificial time-series spectra, have proven that ‘constrained genetic disentangling’ is performing correctly and efficiently, albeit with high demand on CPU time. Since genetic optimization can be easily parallelized, we expect our second release to run on cluster in a less time-consuming way.