Joint Survey Processing. III. Compact Oddballs in the COSMOS Field -- Little Red Dots and Transients

Yu-Heng Lin, Andreas L. Faisst, Ranga-Ram Chary, Anton M. Koekemoer, Joseph Masiero, Daniel Masters, Vihang Mehta, Harry I. Teplitz, Gregory L. Walth, John R. Weaver

First listed 2025-04-09 | Last updated 2025-04-09

Abstract

We present the HST ACS G800L grism spectroscopy observation of the faint active galactic nuclei (AGN) candidates in the COSMOS field at redshift of 6 selected by the point-source morphology and the photometry drop-off at 8000Å. Among the sample of 7 objects, only one is detected by multiple bands, and has similar shape of spectral energy distribution as the so-called ``little red dots'' JWST selected AGN candidates, but our object is 3 magnitude brighter than the JWST sample. We draw the upper limit of the AGN luminosity function $Φ=1.1\times 10^{-7}$Mpc$^3$ mag$^{-1}$ for $M_{UV}$=$-21$ at redshift of 6. The rest of the sample shows inconsistent flux density when comparing magnitudes of HST ACS F814W to the Subaru $i$-band and $z$-band magnitudes combined. The HST ACS G800L grism observation shows that this inconsistency cannot be created from an emission line. Therefore, we speculate that these objects are transients with the light curve decay timescale at most 6 years in observed frame.

Short digest

HST/ACS G800L grism follow-up of seven compact, 8000 Å–dropout COSMOS point sources tests the faint z≈6 AGN candidate pool. Only one object (772319) is detected in multiple bands with a V-shaped, LRD-like SED and is ~3 mag brighter than typical JWST LRDs; no Lyα or other common lines appear in the grism. From the non-detections they place an upper limit on the AGN UV luminosity function of Φ=1.1×10^-7 Mpc^-3 mag^-1 at M_UV=−21. The remaining sources show F814W vs Subaru i+z flux mismatches not attributable to emission lines, implying long-lived transients with observed-frame decay timescales ≤6 years and highlighting transient contamination of faint-AGN searches.

Key figures to inspect

• Figure 1: Compare 772319 (top) and 342154 (bottom) image cutouts across HST and ground/JWST bands to see that 772319 is multiband-detected while 342154 remains undetected in deep NIRCam/MIRI, supporting a transient or extreme SED interpretation for the faint subset.
• Figure 2: Two-PA stacked 2D grism spectra for 772319; inspect the marked Lyα position and the absence of any coincident features in both PAs to verify that an emission line cannot explain the F814W excess.
• Figure 3: EW(Lyα) versus M_UV with S/N=3 curves; read off the Lyα EW upper limits consistent with the F814W flux density and compare to AGN/SFG loci to see that any Lyα would be too weak to drive the photometric break.
• Figure 4: SED of 772319 contrasted with a cool dwarf template and average LRD SEDs; note the V-shape, brightness offset (~2.6–3.5 mag brighter than JWST-selected LRDs), and location relative to LRD color cuts to assess its status as a bright LRD analog.