2025A&A...704A..61M

Context. Massive amounts of spectroscopic data obtained by stellar surveys are feeding an ongoing revolution in our knowledge of stellar and Galactic astrophysics. Analysing these datasets to extract the best possible astrophysical parameters on short timescales is a considerable challenge.
Aims. The differential analysis method is known to return the most precise results in the spectroscopic analyses of F-, G-, and K-type stars; however, it can only be applied to stars with similar parameters. Our goal is to present a procedure that significantly simplifies the identification of spectra from stars with similar atmospheric parameters within extensive spectral datasets. This approach allows for a quick application of differential analyses in these samples, thereby enhancing the precision of the results.
Methods. We used projection maps created by the t-SNE dimensionality reduction algorithm applied directly to the spectra using pixels as dimensions. To test the method, we used more than 7300 high-resolution UVES spectra of about 3000 stars in the field of view towards open and globular clusters. As our reference, we used 1244 spectra of 274 stars with well-determined and high-quality atmospheric parameters.
Results. We calibrated a spectral similarity metric that allowed us to identify stars in a t-SNE projection map with parameters that differed by ±200 K, ±0.3 dex, along with ±0.2 dex in effective temperatures, surface gravities, and metallicities, respectively. We achieved completeness between 74-98% and typical purity between 39-54% in this selection. With these data in hand, we have the ability to fully enable the detection of stars with similar spectra for a successful differential analysis. In this work, we apply this method to evaluate the accuracy and precision of four atmospheric parameter catalogues, identifying regions of the parameter space where the spectral analysis methods need improvement.