A White Dwarf Catalog from LAMOST DR11 Using Deep Learning

Abstract White dwarfs, also referred to as degenerate dwarfs, represent the ultimate fate for the majority of stars in the Universe, unless their mass is large enough for them to evolve into black holes or neutron stars. In recent years, the astronomical community has experienced an unprecedented influx of data from initiatives like LAMOST, the Sloan Digital Sky Survey, and Gaia. This surge in observational data has underscored the pressing need for efficient analysis and categorization of newly discovered white dwarfs. Traditional methods for star classification heavily rely on manual visual identification, which is a laborious and time-consuming process. This study presents a pioneering approach by integrating 1D spectral data and 2D spectral images to enhance the accuracy of white dwarf classification. It involves the development of convolutional neural networks for processing 1D spectral data, alongside residual neural networks for handling 2D spectral images. The synergy between these methodologies significantly improves the accuracy of white dwarf classification. Our model represents a substantial advancement in white dwarf classification accuracy, alleviating the substantial manual workload and time required for classification, and aligning with the requirements of contemporary astronomy characterized by the era of big data. Our two-class classification model with multimodal data achieved an accuracy rate of 98.69% on the test set. Subsequently, we design a white dwarf identification flowchart applied for the LAMOST Data Release 11 low-resolution spectra. Finally, we compile a comprehensive white dwarf catalog, which includes 362 new white dwarfs, alongside 4456 known white dwarfs.