Spectangular - a software tool for spectral disentangling

The software Spectangular is for spectral disentangling via singular value decomposition with global optimization of the orbital parameters of the stellar system or radial velocities of the individual observations.

Binary stars are of outstanding importance for stellar physics, since the masses of both components can be measured by combining spectroscopy and photometry and/or interferometry.  The mass and the chemical composition defining the evolutionary state and the stellar structure  (Vogt-Russell Theorem) in case that rotation and magnetic fields can be neglected. Hence, with data on spectroscopic binaries which are eclipsing as well, it is possible to study models on stellar evolution in detail. For such detailed analysis it is indispensable to use data with high quality and to separate the spectra of both components. For this purpose the tool Spectangular was written. It performs the technique of spectral disentangling on a time-series of spectra. Spectangular works with data in the wavelength scale which need to be re-sampled to a logarithmic scale. The individual spectra need to be spread homogeneously over  the orbital period of the system and have to be of equal quality in the sense of normalization and signal- to noise ratio. The code solves via singular value decomposition (SVD) for the two spectra  of the components. Additionally, the SVD is coupled to a global optimization algorithm (Downhill Simplex) to optimize the radial velocities of orbital parameters of the system. Hence, the tool outputs the component spectra and system parameters. Performing SVD in the wavelength domain has some advantageous over Fourier codes. However, the size of required RAM is high and the claims on  matrix classes as well. Hence, only smaller wavelengths ranges can be treated in one run. Furthermore, the system to be solved via SVD needs to be overdetermined. This requires at least two spectra for SB1 and three spectra for SB2 systems. As more spectra at different phase positions are used, as smaller will be the noise in the results. Sufficient data will result in spectra with a lower noise level as the individual spectra.


The two component spectra are disentangled from a set of 18 spectra of the binary Capella. The primary (red) is an evolved slow rotating (vsini~4km/s) star while the secondary is moving  through the Herzsprung gap. Lithium, as an indicator for evolution, is still prominent in the secondary but is much less abundant in the primary. This can be seen from the Lithium line at 6708 A.

We also used this code for spectral classification of a Wolf-Rayet-Star. Since these stars are variable due to their wind and a possible wind-wind colliding region, it is a challenging task for spectral disentangling. However, it was possible to separate the two spectra of the system R145. See Fig. 6 in  arXiv:1610.07614

The paper on the code can be found here: Sablowski & Weber 2017 A&A 597, A125