A contact binary is a heavily interacting binary system composed of two component stars filled with Roche lobes and surrounded by a shared envelope. With the publication of hundreds of light curves of contact binaries, existing approaches often take many hours or days to extract the parameters of contact binaries.
Dr. Ding Xu and Prof. Ji Kaifan of the Chinese Academy of Sciences Yunnan Observatories (CAS), in conjunction with Li Xuzhi of China's University of Science and Technology, have suggested a machine learning-based approach for swiftly obtaining the parameters and errors of a contact binary. On October 18, this report was published in The Astronomical Journal.
The researchers initially employed a neural network (NN) to construct a mapping link between the characteristics of the contact binary stars and the light curves, resulting in one model without the effect of the third light and one model with the influence of the third light. The inaccuracy of the light curves created by these two models is less than one-thousandth of a magnitude, and the contact binaries' parameters and accompanying errors may be derived fast by combining the Markov chain Monte Carlo approach (MCMC). When compared to standard approaches, our solution not only satisfies the accuracy criteria but significantly improves speed by four orders of magnitude under the same running conditions.
This approach may be used to calculate the parameters of a large number of a contact binary. The researchers will then undertake a statistical analysis of the contact binaries in the space telescope's TESS survey data and the ground telescope's ZTF survey data.
Journal Information: Xu Ding et al, Fast Derivation of Contact Binary Parameters for Large Photometric Surveys, The Astronomical Journal (2022). DOI: 10.3847/1538-3881/ac8e66
Comments