J-HERTz: J-PLUS Heritage Exploration of Radio Targets at z $<$ 5

TL;DR

J-HERTz delivers a public, cross-matched catalog combining J-PLUS optical narrow-band photometry, LoTSS DR2 radio data, and WISE infrared measurements over roughly $2100\,\mathrm{deg^2}$, enabling probabilistic classifications for ~${\sim}4.9\times10^5$ sources. It leverages Bayesian neural networks (BANNJOS) to assign galaxies, QSOs, and stars with PDFs, and introduces a conflated photo-z that improves redshift accuracy for ~${\sim}2.35\times10^5$ galaxies, complemented by robust spectroscopic redshifts where available. A novel radio-loudness proxy based on the low-frequency radio-to-IR flux ratio $\log(f_{144\mathrm{MHz}}/f_{\mathrm{W2}})$ reveals a bimodal QSO distribution and a substantial population of optically quiescent, radio-loud galaxies, including relic jet systems, all linked to host properties like stellar mass and sSFR. The catalog also identifies 831 candidate Galactic radio stars and demonstrates the value of radio–optical–IR synergies for studying AGN life cycles, jet feedback, and the co-evolution of galaxies and their central engines, with future expansion anticipated from J-PAS data.

Abstract

We introduce J-HERTz (J-PLUS Heritage Exploration of Radio Targets at $z < 5$), a new multi-wavelength catalog that combines optical narrow-band photometry from J-PLUS, infrared observations from WISE, and deep low-frequency radio data from LoTSS for nearly half a million sources across 2,100 deg$^2$ of the northern sky. Key innovations of J-HERTz include Bayesian neural network classifications for 390,000 galaxies, 31,000 quasars, and 20,000 stars, along with significantly improved photometric redshifts for 235,000 galaxies compared to previous J-PLUS DR3 and LoTSS DR2 estimates. We identify 831 candidate Galactic radio stars, which, if confirmed, would constitute a significant addition to the number of radio-emitting stars identified to date. Among radio-loud galaxies with spectroscopic observations, $\gtrsim$20% lack Seyfert or LINER signatures, indicating a substantial population of optically quiescent radio galaxies, in agreement with previous works. Spectral energy distribution fitting of their host galaxies using J-PLUS photospectra reveals systematically low specific star formation rates, consistent with quenched stellar populations. J-HERTz thus provides a powerful dataset to exploit radio-optical synergies, enabling studies that span from the origin of stellar radio emission to the AGN life cycle and the role of jet activity in shaping host galaxy evolution.

Figures (15)

• Figure 1: Sky coverage of the 2100 deg$^2$ common area of J-PLUS DR3 (green shaded area) and LoTSS DR2 (blue shaded area).
• Figure 2: Three objects included in J-HERTz, with their unique J-PLUS identification and their SDSS DR18 Almeida2023 spectroscopic redshifts. Left panels: optical image from J-PLUS in gray scale, with the J-PLUS and LoTSS positions (blue square and red dot respectively), and the radio contours (black lines) from LoTSS drawn at 1, 2, 3, 5 and 7$\sigma$. The blue circle is the J-PLUS aperture selected to match the size of the SDSS fiber in each case (either 2 arcsec for BOSS or 3 arcsec for SDSS). The 1 arcsec and 6 arcsec lines show the resolutions of J-PLUS and LoTSS respectively, when visible. Right panels: Fluxes and error bars measured in the 12 J-PLUS filters with the specified aperture as the colored symbols (squares for broad and circles for narrow bands, most error bars are smaller than the symbol size) and their SDSS DR18 spectra in gray, smoothed and re-scaled to the J-PLUS flux in the $r$ band.
• Figure 3: Mean probabilities of being a galaxy, QSO or star for all objects in J-HERTz according to the BANNJOS delPino2024 classification. Each voxel is colored according to the number of points inside. The voxels inside the blue, orange and green contours contain objects classified as galaxies, QSOs and stars with a mean probability above 0.9, which correspond to over 90% of the catalog and are the ones we consider for the rest of this work.
• Figure 4: Comparison between different photo-z values in J-PLUS (left), LoTSS Duncan2022, and our combined estimate from J-PLUS and LoTSS conflation (right), with the spectroscopic z taken from SDSS, DESI or HETDEX, when available, for sources classified as galaxies with both photometric and spectroscopic z $<$ 1. The solid black line represents the one-to-one relation, and the dashed lines are drawn at $\pm 3\sigma_{\text{MAD}}$ (the sigma is indicated at the top of each panel). The conflation improves the accuracy and precision of both original photo-zs.
• Figure 5: Left: LoTSS stellar masses taken from Hardcastle2023 adjusted for our best_z against J-PLUS stellar masses computed by CIGALE Arizo2025 for galaxies with r $<$ 21 and low uncertainty (see text). The orange contours contain 90% and 50% of galaxies with r mag above 20 and redshift above 0.2 (faint), while the blue contours contain 90% and 50% of galaxies with r mag below 20 and redshift below 0.2 (bright). Right: Distribution of the J-PLUS stellar mass of galaxies with r $<$ 21 and low uncertainty. The most extreme outliers were adjusted with redshifts that differed considerably from the original J-PLUS photo-z and should not be trusted.