MIDIS: The identification of deep MIRI-red sources as candidates for extreme Balmer-break and line emitting galaxies at high-z

I. Jermann, G. Brammer, S. Gillman, T. R. Greve, L. A. Boogaard, J. Melinder, R. A. Meyer, P. G. Pérez-González, P. Rinaldi, J. L. Colina, G. Östlin, G. Wright, J. Álvarez-Márquez, A. Bik, K. I. Caputi, A. Crespo Gómez, L. Costantin, J. Hjorth, E. Iani, S. Kendrew, A. Labiano, D. Langeroodi, F. Peissker, C. Prieto-Jiménez, J. P. Pye, T. V. Tikkanen, F. Walter, P. van der Werf, T. Henning, M. Shuntov

First listed 2026-02-19 | Last updated 2026-02-19

Abstract

We investigate the detection and nature of 5.6~μm MIRI-red sources in the MIRI Deep Imaging Survey (MIDIS), covering 2.4~arcmin$^2$ in the Hubble Ultra Deep Field. MIDIS is the deepest JWST/MIRI survey to date, probing faint limits and enabling studies of rare high-redshift galaxy populations. We define MIRI-red sources as those detected at 5$σ$ significance in MIRI/F560W with red colors: $m_{\rm F444W} - m_{\rm F560W} \ge 0.5$. Using an empirical methodology, we estimate the purity and completeness of MIRI detections and find that a 5-sigma detection at 28.75 mag has a purity of 92\% and completeness of 54\%. We identify seven MIRI-red galaxy candidates, including an F115W dropout consistent with a high-redshift galaxy candidate. We explore possible physical origins for the MIRI-red population, including active galactic nuclei, dust-obscured galaxies, extreme emission-line galaxies, evolved stellar populations, and Little Red Dots (LRDs). Given the proximity of the F444W and F560W filters and the depth of MIDIS, MIRI-red galaxies are consistent with emission-line galaxies with $EW_0(Hα) \ge 750$ Å or $EW_0(Hβ+ [OIII]) \ge 600$ Å, or high-redshift Balmer breaks of at least 1.6. We also discuss an extreme MIRI-red galaxy undetected in F444W, a potential MIRI-only source, for which we derive $EW_0(Hα) \sim 6000$ Å and $EW_0(Hβ+ [OIII]) \sim 4000$ Å, or high-$z$ LRD analogs with Balmer breaks of 6.3. Finally, we find fewer MIRI-red detections than expected from extrapolations of the H$α$ or H$β$+[OIII] line luminosity functions, consistent with previous deep searches, while the absence of $z>10$ LRD candidates agrees with theoretical expectations for the MIDIS volume.

Short digest

Using the deepest 5.6 μm MIRI imaging of the HUDF (2.4 arcmin²), MIDIS defines “MIRI‑red” sources as 5σ F560W detections with mF444W−mF560W ≥ 0.5 and empirically calibrates the background, finding pipeline S/N is overestimated and achieving 92% purity and 54% completeness at 28.75 mag. Seven MIRI‑red candidates are found (including an F115W dropout), with colors consistent with extreme rest‑optical line emitters requiring EW0(Hα) ≥ 750 Å or EW0(Hβ+[OIII]) ≥ 600 Å, or with high‑z Balmer breaks ≥1.6. An extreme MIRI‑only candidate undetected in F444W implies EW0(Hα) ∼ 6000 Å and EW0(Hβ+[OIII]) ∼ 4000 Å, or an LRD‑like Balmer break ≈6.3. Fewer detections than extrapolated from Hα/Hβ+[OIII] luminosity functions and the absence of z>10 LRDs underscore the rarity of such systems and the value of mid‑IR selection for isolating the most extreme cases.

Key figures to inspect

• Figure 1: Use the field mosaic and Deep/Shallow contours to see where the seven MIRI‑red candidates (magenta squares) sit relative to the deepest F560W coverage and to ALMA Band 3/6 contours—check any positional coincidences that would favor dusty/AGN interpretations.
• Figure 2: Inspect the bright‑source mask; it leaves 91.3% of the MIDIS Deep area usable—verify that candidates avoid masked edges where background estimation and color selection could be biased.
• Figure 3: Background noise versus aperture size compares empty‑aperture scatter (black) to variance‑map errors before/after rescaling (red); this quantifies the S/N overestimation and justifies the empirically rescaled photometric uncertainties used for the selection.
• Figure 4: Purity and completeness curves versus F560W magnitude/flux; at 28.75 mag the survey reaches 92% purity and 54% completeness—use this to gauge the reliability of the seven candidates and the expected number missed at fainter levels.