Autonomous Disentangling for Spectroscopic Surveys (2405.19391v2)

Abstract: A suite of spectroscopic surveys is producing vast sets of stellar spectra with the goal of advancing stellar physics and Galactic evolution by determining their basic physical properties. A substantial fraction of these stars are in binary systems, but almost all large-survey modeling pipelines treat them as single stars. For sets of multi-epoch spectra, spectral disentangling is a powerful technique to recover or constrain the individual components' spectra of a multiple system. So far, this approach has focused on small samples or individual objects, usually with high resolution ($R \gtrsim 10.000$) spectra and many epochs ($\gtrsim 8$). Here, we present a disentangling implementation that accounts for several aspects of few-epoch spectra from large surveys: that vast sample sizes require automatic determination of starting guesses; that some of the most extensive spectroscopic surveys have a resolution of only $\approx 2,000$; that few epochs preclude unique orbit fitting; that one needs effective regularisation of the disentangled solution to ensure resulting spectra are smooth. We describe the implementation of this code and show with simulated spectra how well spectral recovery can work for hot and cool stars at $R \approx 2000$. Moreover, we verify the code on two established binary systems, the Unicorn'' andGiraffe''. This code can serve to explore new regimes in survey disentangling in search of massive stars with massive dark companions, e.g. the $\gtrsim 200,000$ hot stars of the SDSS-V survey.