Medan, Ilija; Way, Zachary; Rojas-Ayala, B谩rbara; Stringfellow, Guy S.; Sayres, Conor; Stassun, Keivan G.; Casey, Andrew R.; L茅pine, S茅bastien; Galligan, Emma; Souto, Diogo; Saydjari, Andrew K. (2025).听.听Astronomical Journal, 170(6), 302.听
The Sloan Digital Sky Survey V (SDSS-V) will collect a large number of spectra鈥攄etailed 鈥渇ingerprints鈥� of light鈥攆rom听M dwarfs, which are small, cool stars that are very common in our galaxy. However, analyzing these stars is challenging because their atmospheres produce complex spectra filled with many overlapping听molecular absorption features听(dark bands where molecules absorb light), making it hard to determine key properties like temperature or composition using traditional models. To get around this, researchers often use听machine learning, training algorithms to estimate stellar properties by learning from higher-quality data. But these methods can introduce errors, partly because of how the spectra are听normalized鈥攁 process that adjusts the data to make comparisons easier.
In most stars, normalization involves identifying a smooth baseline (the听continuum) and measuring how much light is absorbed relative to it. For M dwarfs, this is difficult because their spectra are so dominated by absorption features that a clear baseline is hard to define. To solve this, the authors developed a new method that estimates a听鈥减蝉别耻诲辞肠辞苍迟颈苍耻耻尘鈥�鈥攁n approximate baseline鈥攂y identifying the least absorbed parts of the spectrum and fitting a smooth curve through them. They tested and refined this approach using simulated data designed to mimic real observations, including effects from instruments and noise.
The results show that this new method produces more consistent spectra for stars with similar properties and better distinguishes between different types of M dwarfs compared to existing techniques. This improvement is important for making more accurate measurements of stellar properties in large surveys like SDSS-V, especially when using advanced modeling approaches such as machine learning.
Figure 1.听H-R diagram of the number of spectra observed with BOSS for the Milky Way Mapper (MWM) during SDSS-V until MJD = 60715 (2025 February 9). The top and right histograms show the cumulative distributions of in听G 鈥�听RP and听MG, respectively. For the听MG听distribution, the location of main-sequence spectral types as a function of听MG听from M. J. Pecaut & E. E. Mamajek () are shown for reference. From the听MG听cumulative distribution on the right, an impressive 鈭紀ne-third of BOSS MWM spectra are of M dwarfs.