A new code for low-resolution spectral identification of white dwarf binary candidates

Abstract

Close white dwarf binaries (CWDBs) are considered to be progenitors of several exotic astronomical phenomena (e.g., type Ia supernovae, cataclysmic variables). These violent events are broadly used in studies of general relativity and cosmology. However, obtaining precise stellar parameter measurements for both components of CWDBs is a challenging task given their low luminosities, swift time variation, and complex orbits. High-resolution spectra (R$> 20 000$) are preferred but expensive, resulting in a sample size that is insufficient for robust population study. To release the full potential of the less expensive low-resolution spectroscopic surveys, and thus greatly expand the CWDB sample size, it is necessary to develop a robust pipeline for spectra decomposition and analysis. We used an artificial neural network (ANN) to build spectrum generators for DA/DB white dwarfs and main-sequence stars. The best-fit stellar parameters were obtained by finding the least $\chi^2$ solution to these feature lines and the continuum simultaneously. We demonstrate the reliability of our code with two well-studied CWDBs, WD 1534+503 and PG 1224+309. We also estimate the stellar parameters of 14 newly identified CWDB candidates, most of which are fitted with double component models for the first time. Our estimates agree with previous results for the common stars and follow the statistical distribution in the literature. The application of our code to a large volume of white dwarf binary candidates will offer important statistic samples to stellar evolution studies and future gravitational wave monitoring.