Taxonomy of 14042 asteroids from Gaia DR3 reflectance spectra

Asteroid reflectance spectra provide key constraints on surface composition. Data Release 3 (DR3) enables the study of asteroids through near-ultraviolet (NUV) to visible reflectance spectra. Gaia Our aim is to classify asteroids using DR3 spectra and provide a homogeneous taxonomical framework. Owing to systematics affecting DR3 data, direct comparison with previous taxonomies has to be performed with caution; therefore, we developed a classification scheme tailored to and linked the resulting taxa to established classes. Gaia Gaia Gaia We selected the highest-quality spectra using DR3 quality flags and applied uncertainty thresholds to mitigate spectral artifacts, retaining over one-third of the original sample at the least noisy wavelengths. To improve compositional discrimination, we included albedo information, thus reducing the final sample to about one-fourth of its initial size. We then iteratively applied dimensionality reduction and clustering techniques to identify the spectral taxa. Gaia We classified asteroids into 13 taxonomic classes: A, B, C, D, E, F, G, K, L, M, P, S, and V, representing an increase in compared to the number of objects classified in previous spectral classifications. The largest relative increase was found for the K class. The inclusion of NUV wavelengths allowed the separation of B and F types within the C complex and facilitated the identification of G types. The dynamical distribution follows expected trends: S types dominate the inner and middle main belt; C-complex asteroids are prevalent in the outer main belt; and D types are prevalent beyond the main belt. ̊atioclassif We present a taxonomical classification of asteroids based on DR3 reflectance spectra. NUV coverage is critical for disentangling primitive classes within the C complex. Although artifacts in DR3 require caution when one is comparing median spectra with other datasets, this classification provides a robust reference for future releases with larger observed samples.