arXiv subjects

Caterina Bracci

Publications and source records attributed to Caterina Bracci.

5 recordsLinked to original sources

arXiv2026

Unsupervised selection and characterisation of Little Red Dots in JWST surveys with manifold learning

Little Red Dots (LRDs) are compact, red sources discovered at high redshift by JWST whose physical nature and selection function remain debated. We investigate whether an unsupervised machine-learning approach applied to multi-band photometry can identify LRD-like objects, and other populations, without relying on predefined colour cuts. Using UMAP, a manifold-learning (dimensionality-reduction) method, we place ~242,000 isolated, well-measured sources from the ASTRODEEP-JWST catalogue on a two-dimensional map, where objects with similar broadband colours, morphology, and photometric redshift lie close together. We then use spectroscopically confirmed LRDs to identify where LRD-like objects lie within this map, compare the resulting areas with published colour cuts, and validate our data-driven selection against archival NIRSpec spectra from the DJA. We find that the spectroscopically selected LRDs concentrate in two well-defined regions with no colour cut imposed, tracing populations that differ mainly in redshift, a difference imprinted in their broadband colours. The main region reaches a purity of ~0.78 at ~0.82 completeness on the spectroscopically classified subset, competitive with, or cleaner than, literature colour cuts, and yields ~100 additional candidates. We also test the method as a general tool for population discovery: the manifold recovers the locations of brown dwarfs and broad-line AGN with no explicit criterion, and isolates rare pathological outliers. Overall, unsupervised manifolds, anchored by sparse high-confidence spectroscopic labels, provide an efficient, assumption-light framework for characterising populations, comparing selection methods on a common basis, and discovering rare objects in large photometric datasets.

astro-ph.GA↗

arXiv2026

Milky-Way-like stars in a galaxy core 8 billion years ago revealed by gravitational lensing

The assembly of stellar-dominated cores in elliptical galaxies is key to understanding how cosmic structures evolved. Gravitational lensing offers unique insights into the nature of their stars. We report the discovery of the smallest known quadruply lensed quasar (radius ~0.2"), whose lensing galaxy at redshift 1.055 (5.5 billion years after the Big Bang) features a lensing mass of only ~2x10^10 M_sun. A Bayesian analysis, based on the system's exceptional properties and standard scaling relations, allowed us to sample the central galactic initial mass function with unmatched accuracy and in a previously uncharted regime in terms of mass and redshift. We found it consistent with the Milky Way one, while excluding bottom-heavy functions. This suggests that the core either grew slowly or underwent early disruptive events altering its stellar build-up, in contrast with the classical view that bulges form rapidly and remain unchanged by later interactions.

astro-ph.GA↗

arXiv2026

One cloud is not enough: extreme conditions bias chemical abundances in high-redshift galaxies

Since its launch, JWST has opened an unprecedented opportunity to characterise the ionised ISM of high-redshift galaxies using well-established rest-frame UV/optical diagnostics from the local Universe. At the same time, these observations challenge the validity of such classical methods when applied to the extreme environments typical at high redshift. We present an in-depth analysis of the ISM in three representative case studies at $z=2 - 6$ (MARTA 4327, the Sunburst Arc and RXCJ2248-ID) conducted within a multi-cloud photoionisation modelling framework (HOMERUN). We show that even a small fraction of unresolved high-density clumps can contribute more than half of the observed flux of auroral lines, while only negligibly to standard optical density tracers. As a result, $T_{\mathrm{e}}$-method metallicities can be underestimated by $\sim 0.15 - 0.3$ dex, as for MARTA 4327. By modelling rest-frame UV and optical data, we demonstrate that discrepancies between abundances obtained from diagnostics tracing different zones do not necessarily imply chemical inhomogeneities. In RXCJ2248-ID, the disagreement between UV and optical N/O may naturally arise from ionisation and density structure alone. In contrast, we find evidence for genuine chemical stratification in the Sunburst Arc, where a component enriched in nitrogen coexists with a chemically normal one. Finally, we argue that very-high-ionisation lines may be explained within a pure star-formation scenario invoking matter-bounded regions. However, in the case of RXCJ2248-ID, we cannot rule out a minor contribution from an AGN based solely on the observed fluxes. These results indicate that classical diagnostics can be significantly biased in high-redshift galaxies and that self-consistent, physically motivated tools are therefore essential to properly interpret the complex ISM conditions and chemical enrichment in the early Universe.

astro-ph.GA↗

arXiv2025

Classifying spectra of emission-line regions with neural networks -- An application to integral field spectroscopic data of M33

Emission-line regions are key to understanding the properties of galaxies, as they trace the exchange of matter and energy between stars and the interstellar medium (ISM). In nearby galaxies, individual nebulae can be identified as HII regions, planetary nebulae (PNe), supernova remnants (SNR), and diffuse ionised gas (DIG) with criteria on single or multiple emission-line ratios. However, these methods are limited by rigid classification boundaries, the narrow scope of information they are based upon, and the inability to account for line-of-sight nebular superpositions. In this work, we use artificial neural networks to classify these regions using their optical spectra. Our training set consists of simulated spectra, obtained from photoionisation and shock models, and processed to match observations obtained with MUSE. We evaluate the performance of the network on simulated spectra for a range of signal-to-noise (S/N) levels and dust extinction, and the superposition of different nebulae along the line of sight. At infinite S/N the network achieves perfect predictive performance, while as the S/N decreases, the classification accuracy declines, reaching an average of ~80% at S/N(H$α$)=20. We apply our model to real spectra from MUSE observations of the galaxy M33, where it provides a robust classification of individual spaxels, even at low S/N, identifying HII regions and PNe and distinguishing them from SNRs and diffuse ionized gas, while identifying overlapping nebulae. We then compare the network's classification with traditional diagnostics and find satisfactory agreement. Using activation maximisation maps, we find that at high S/N the model mainly relies on weak lines (e.g. auroral lines of metal ions and He recombination lines), while at the S/N level typical of our dataset the model effectively emulates traditional diagnostic methods by leveraging strong nebular lines.

astro-ph.GA↗

arXiv2024

Datacube segmentation via Deep Spectral Clustering

Extended Vision techniques are ubiquitous in physics. However, the data cubes steaming from such analysis often pose a challenge in their interpretation, due to the intrinsic difficulty in discerning the relevant information from the spectra composing the data cube. Furthermore, the huge dimensionality of data cube spectra poses a complex task in its statistical interpretation; nevertheless, this complexity contains a massive amount of statistical information that can be exploited in an unsupervised manner to outline some essential properties of the case study at hand, e.g.~it is possible to obtain an image segmentation via (deep) clustering of data-cube's spectra, performed in a suitably defined low-dimensional embedding space. To tackle this topic, we explore the possibility of applying unsupervised clustering methods in encoded space, i.e. perform deep clustering on the spectral properties of datacube pixels. A statistical dimensional reduction is performed by an ad hoc trained (Variational) AutoEncoder, in charge of mapping spectra into lower dimensional metric spaces, while the clustering process is performed by a (learnable) iterative K-Means clustering algorithm. We apply this technique to two different use cases, of different physical origins: a set of Macro mapping X-Ray Fluorescence (MA-XRF) synthetic data on pictorial artworks, and a dataset of simulated astrophysical observations.

cs.LG↗