arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 2604.03182

DeepDISC-Euclid: Source Classification and Photometric Redshifts in Euclid Deep Field North With a Pixel-Level Deep Learning Approach

Yuanzhe Jiang·Yue Shen·Grant Merz·Shurui Lin·Xin Liu·Zhiwei Pan·Mingyang Zhuang·William Roster·Mara Salvato·Malgorzata Siudek·Grant Stevens
View original source ↗Read source document ↗Export metadata ↓

Abstract

The first Euclid Quick Data Release (Q1) provides extensive imaging and spectroscopic data for hundreds of millions of photometric objects across several deep fields. Accurate classifications and photometric redshifts (photo-z) for these sources are crucial to maximizing the value of these data. In this work, we perform source classification and photo-z estimation for the Euclid Deep Field North (EDF-N) around the North Ecliptic Pole, using a deep learning framework (DeepDISC) that learns and infers using 9-band images simultaneously. We train three dedicated models for (1) source detection and classification, (2) galaxy photo-z, and (3) quasar photo-z. The Euclid Q1 input source catalog, and classifications and spectroscopic redshifts (spec-z) from the Dark Energy Spectroscopic Instrument Data Release 1 are adopted as our training data. DeepDISC source detection achieves overall completeness of ~93% and purity of ~80% if using the Euclid source catalog as the ground truth. Using a JWST source catalog within EDF-N as the reference, we estimate a true purity of ~ 90% for DeepDISC sources. About 99.2%, 99.0%, and 84.8% of stars, galaxies, and quasars, respectively, are correctly recovered with their spectroscopic classifications. The DeepDISC photo-zs show good agreement with spectroscopic redshifts, for both galaxies and quasars. Comparisons with other Euclid Q1 products demonstrate that DeepDISC provides comparable or improved performance in source detection/deblending, classification and photo-z, especially for quasars. These results demonstrate the potential of pixel-level deep learning approaches for large-scale sky surveys such as Euclid and Roman, which will continue to improve with better training labels. We release the full DeepDISC source catalog (~13 million objects) for EDF-N with classifications and photo-zs, including photo-z probability distributions.

Explore related subjects

astro-ph.GAastro-ph.IM

Keep this discovery

BibTeXRIS

Yuanzhe Jiang, Yue Shen, Grant Merz, Shurui Lin, Xin Liu, Zhiwei Pan, Mingyang Zhuang, William Roster, Mara Salvato, Malgorzata Siudek, Grant Stevens. 2026-04-03. DeepDISC-Euclid: Source Classification and Photometric Redshifts in Euclid Deep Field North With a Pixel-Level Deep Learning Approach. https://arxiv.org/abs/2604.03182

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Fast Dynamical Modelling of Milky Way Globular Clusters -- II. Impacts of Black Hole Prescriptions

The populations of stellar-mass black holes (BHs) in globular clusters (GCs) play a key role in their dynamical evolution, however the mechanisms surrounding their formation and retention are uncertain. In this work, we extend the analysis of Paper I by fitting coupled rapid cluster evolution and multimass equilibrium models to a large sample of Milky Way GCs, under a variety of prescriptions for stellar evolution, BH formation and supernovae (SN) natal kicks. We explore the impacts of adopting SSE or PARSEC (through SEVN) prescriptions for BH initial-final mass relations, the rapid or delayed SN fallback mechanisms, and an ad hoc grid of kick strengths ejecting between 40 and 80 per cent of all BHs formed. All models reproduce the same present-day conditions despite starting from notably different initial BH populations, due to the correlation found between the initial cluster densities and initial BH mass fractions. A linear relationship is found between the (log) initial half-mass density and the initial BH mass fraction, with the SEVN models resulting in median densities ($\rho_{h,0} \sim 10^{7.2\pm1.1}\,{M_\odot pc^{-3}}$) nearly an order of magnitude higher than those of SSE ($\rho_{h,0} \sim 10^{6.4\pm0.9}\,{M_\odot pc^{-3}}$). We also find that both the bottom-light initial mass functions and the present-day BH mass fractions previously inferred are relatively robust against the stellar evolution models and natal kick prescriptions assumed. Finally, we discuss the implications of these results on the expected numbers and properties of dynamical binary-BH mergers, and the growth of intermediate-mass BHs.

astro-ph.GA↗
arXiv2026

SPURS: An Ultra-deep View Inside the Compact, Nitrogen-Enriched Nuclei of Little Red Dots

We present the first ultra-deep rest-UV spectroscopy of four UV-bright Little Red Dots (LRDs), obtained from the SPURS Cycle 4 Large Program. The spectra reveal broad CIV (FWHM $\approx2700-2800$ km s$^{-1}$) in two LRDs, alongside narrow-line densities elevated above star-forming galaxies ($n_e\sim10^4-10^5$ cm$^{-3}$, reaching $10^6$ cm$^{-3}$ in the most extreme source) and nitrogen-enhancements in all four LRDs. We detect broad HeII emission (FWHM $\approx930$ km s$^{-1}$) in one LRD, and two others with fast P-Cygni absorption ($\gtrsim2200$ km s$^{-1}$). Strong interstellar absorption lines and Ly$\alpha$ damping wings reveal the UV continuum is deeply embedded in neutral gas ($N_{\rm HI}\gtrsim10^{22}$ cm$^{-2}$) in all four LRDs. Detections of fluorescent FeII and OI emission and fine-structure absorption indicate this gas lies close to the UV-emitting region. In archival $z>4$ samples, we find nitrogen and strong CIII] emission are significantly more common in LRDs than in the galaxy population. The transmission of broad CIV, tracing the broad-line region or cocoon, depends on rest-optical color within our sample, consistent with an orientation-dependent picture in which bluer, less obscured sightlines offer a more direct, polar view of the central engine and its outflows. We find several potential signatures of very massive stars, whose winds may contribute to nitrogen enhancement. We investigate other abundance patterns expected from supermassive stars but our results are inconclusive. Our results place the UV-emitting region within $\lesssim8$ pc of the LRD nucleus, consistent with an actively assembling nuclear star cluster. Dynamical interactions in this extremely dense environment, including tidal disruption of stars, may explain the high incidence of nitrogen enhancements in LRDs.

astro-ph.GA↗
arXiv2026

Nitrogen-Loud Quasars from the Dark Energy Spectroscopic Instrument. I. Sample Selection and Basic Properties

We present the largest sample to date of nitrogen-loud (N-loud) quasars with strong broad N IV] $\lambda1486$ and/or N III] $\lambda1750$ emission lines over the redshift range $1.6 < z < 4.3$, selected from the Dark Energy Spectroscopic Instrument (DESI) Data Release 1. The final sample contains 1,993 N-loud quasars, corresponding to about 1.2% of the parent quasar sample. The $L_{1450}$ distribution of the N-loud quasars is broadly similar to that of the DESI parent sample, but their redshift distribution is distinct, with a stronger concentration around $z \sim 2.5$--3. Their composite spectrum displays a broadly similar UV continuum shape to that of the parent quasars, while showing significantly enhanced broad nitrogen emission features, including N V, N IV], and N III]. Other metal emission features also show a moderate enhancement. Relative to a control sample matched in redshift and UV continuum luminosity, the N-loud quasars show systematically narrower broad C IV and Mg II emission lines, lower single-epoch virial black hole masses, and higher Eddington ratios, suggesting that N-loud quasars may preferentially appear during a relatively rapid black hole accretion phase. The radio-loud fraction is 10.1%, with the highest fraction among objects exhibiting both N III] and N IV] emission. The catalog provides a statistical baseline for future studies of nitrogen enhancement and its physical origin.

astro-ph.GA↗