Unveiling the Nature of Hot Subdwarfs through Spectroscopy and Machine Learning

Project is funded by Vilnius University Research Promotion Fund.

Project number: MSF-JM-17/2025
Principal investigator: Dr. Carlos Viscasillas Vázquez
Team members: Dr. Markus Ambrosch, Dr. Aidas Medžiūnas, Vladas Šatas 
Main collaborators: prof. Ana Ulla, Dr. Enrique Solano, Dr. Šarūnas Mikolaitis, Dr. Esther Pérez Fernández, PhD student Xabier Pérez-Couto, prof. Minia Manteiga, prof. Laura Magrini, prof. Carlos Dafonte, Dr. Arnas Drazdauskas

 

About the project

The evolutionary pathways of hot subdwarf stars (sds) remain an open question in stellar astrophysics. These compact, helium-burning stars are thought to form primarily through binary evolution, yet their binarity rate is still uncertain. Understanding the role of binarity is crucial, as it directly influences their formation, evolutionary outcomes, and final fate. As in many areas of astrophysical research, addressing this question requires the analysis of vast and highly multidimensional datasets - in particular for this project ~20,000 hot sds candidates from Gaia DR3 - which necessitates developing novel analytical techniques based on Machine Learning (ML), Big Data analytics, and AI. We expect that these new techniques for analyzing hot sds will not only provide deeper insights into their evolution, but also contribute to broader advancements in astrophysical data analysis too. Hence, by leveraging Gaia's unprecedented dataset and advancing ML methodologies, this project will provide a more comprehensive picture of hot sd binarity, its role in stellar evolution, while also driving innovations in ML techniques that can be applied to broader astrophysical research and large-scale data analysis.

Primary objectives

  • Binary Classification - Identify hot sds binary systems within the expanded sample.
  • Companion Characterization - Distinguish between different types of binary companions (e.g., main-sequence stars vs. white dwarfs) by analyzing XP spectra.
  • Spectroscopic Validation - Complement our findings with high-resolution spectroscopy using the Vilnius University Echelle Spectrograph (VUES) instrument of the Molėtai astronomical observatory (MAO) to refine classification criteria.
  • Evolutionary Implications - Asses the statistical properties of the found binary fraction and its impact on current hot sds formation and evolution models.
  • Development of the best performing ML methods for binary classification, companion characterization, and statistical properties analysis.