White Dwarf Spectral Classification Using Gaia XP Spectra: an Unsupervised Machine Learning Approach

• Xabier Perez-Couto ^[1] ; Lara Pallas-Quintela ^[1] ; Minia Manteiga ^[1] ; Eva Villaver ^[2] ; Carlos Dafonte ^[1]
  1. [1] Universidade da Coruña

     Universidade da Coruña

     A Coruña, España

     
  2. [2] Instituto de Astrofísica de Canarias

     Instituto de Astrofísica de Canarias

     Santa Cruz de Tenerife, España

     
• Localización: Proceedings XoveTIC 2024: Impulsando el talento científico / coord. por Manuel Lagos Rodríguez, Tirso Varela Rodeiro, Javier Pereira-Loureiro , Manuel Francisco González Penedo , 2024, págs. 467-470
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • Identifying new white dwarfs (WDs) heavy elements is crucial, as they serve as valuable tools for deducing the chemical characteristics of potential planetary systems accreting material onto their surfaces. To detect metallic WDs, we propose a methodology based on an unsupervised learning technique known as Self-Organizing Maps (SOM). This approach projects a high-dimensional dataset onto a two-dimensional grid, where similar elements are grouped into the same neuron. Using this method, we uncovered 143 bona fide WD candidates in the Gaia space mission with several metallic lines in their spectra, including Ca, Mg, Na, Li, and K. The precision metrics achieved with our method are comparable to those of recent supervised techniques.