Exploring the spectral morphology of ultracool dwarf candidates in Euclid Q1 using k clustering

Several studies have investigated ultracool dwarfs (UCDs) in the Euclid Q1 fields, including compilations of known candidates, spectroscopic confirmation based on spectral indices, and photometric searches for new candidates. Although spectra are available for ∼ 90% of photometric candidates, only a small fraction have spectra of sufficient quality for unambiguous spectroscopic confirmation. The aim of this work is to characterise the spectral morphology of objects that have passed photometric selection in order to assess the information content of low S/N Euclid spectra and evaluate the performance of unsupervised clustering in recovering UCD-like populations and identifying unusual or previously overlooked objects. We applied k clustering to denoised near-infrared spectra of 4217 UCD candidates after excluding severely corrupted spectra and extreme outliers. The resulting partition into 20 clusters was compared with visual inspection and conservative ̧hi^2 template fitting. Four of the 20 clusters exhibit centroid spectra consistent with late-M, L, and early-T dwarf morphologies, and together they contain nearly 1000 objects, corresponding to ∼ 23% of the photometric candidates and ∼ 93% of the candidates identified independently by either visual inspection or spectral-index matching. Several additional clusters display water band-dominated morphologies, which is indicative of cool atmospheres. Examination of individual spectra within the clusters led to the identification of a set of metal-poor subdwarf candidates in the Euclid Q1 data (ten objects spanning the range M7–L4). Unsupervised clustering enables efficient exploration of large heterogeneous spectral datasets and can reveal dominant populations and highlight unusual objects. Although it cannot replace object-by-object confirmation, the approach provides a promising framework for the rapid exploration of the Euclid wide survey.