Infrared-Faint Radio Sources (IFRS) Overview

• Infrared-Faint Radio Sources (IFRS) are extragalactic objects defined by strong 1.4 GHz radio emissions and extreme infrared faintness, indicating high-redshift, AGN-dominated activity.
• Observational studies show IFRSs possess steep spectral indices, compact morphologies, and significant radio polarization, which confirm their non-thermal, AGN-powered nature.
• IFRSs provide a critical probe for early supermassive black hole growth and galaxy evolution, informing selection techniques in comprehensive radio and infrared surveys.

Infrared-Faint Radio Sources (IFRS) are a rare class of extragalactic objects distinguished by their relatively strong radio emission at 1.4 GHz yet extreme faintness or non-detection at near- and mid-infrared wavelengths, specifically at 3.6–3.4 μm. These sources exhibit radio-to-infrared flux density ratios (typically $R≡S_{1.4\,\mathrm{GHz}}/S_{3.6\,\mu\mathrm{m}}$) of order $10^{3}$ or higher, greatly exceeding values seen in starbursts, normal radio-loud AGN, or classical high-redshift radio galaxies (HzRGs). IFRSs have been shown through spectroscopic, radio, and SED analyses to represent predominantly high-redshift ($z\sim2-4$) radio-loud active galactic nuclei (RL AGN), often compact and steep-spectrum, inhabiting massive, frequently dust-obscured host galaxies. Their properties, selection, and cosmological significance are encapsulated below.

1. Definition, Discovery, and Selection Criteria

IFRSs were first identified in the Australia Telescope Large Area Survey (ATLAS) as 1.4 GHz radio sources ($S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$ mJy) with no detectable counterpart in ultra-deep Spitzer 3.6 μm imaging at sensitivities down to $\sim5\,\mu$Jy (Cameron et al., 2011, Norris et al., 2011). This extreme radio-infrared mismatch cannot be accounted for in conventional models of star-forming galaxies or AGN, where synchrotron-bright radio sources at $\geq$ mJy levels always produce detectable IR emission.

The widely adopted, survey-independent selection criteria, introduced by Zinn et al., and employed in most subsequent studies, require:

• A high radio-to-IR flux density ratio:

$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$

• Infrared faintness:

$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$

These cuts reject low-redshift radio-loud AGN, powerful starbursts, and galactic contaminants, isolating a population of extragalactic sources with extreme $R$ ($\sim 500$–$10^{3}$0), often below the IR detection limits in even the deepest imaging (Norris et al., 2011, Herzog et al., 2013, Filipović et al., 2021). Second-generation IFRSs detected via WISE and deep Spitzer surveys are complemented by candidate samples in SERVS, SWIRE, and other fields using similar criteria (Maini et al., 2016). The typical sky density is $10^{3}$1–$10^{3}$2 at $10^{3}$3 mJy (Norris et al., 2011, Zinn et al., 2011).

2. Multiwavelength Properties and Observational Characteristics

Radio: IFRSs span $10^{3}$4 from $10^{3}$5 mJy up to several hundred mJy (Collier et al., 2013, Filipović et al., 2021). The spectra are typically steep: measured indices ($10^{3}$6) show median $10^{3}$7 for classical samples, steeper than the general RL AGN population (median $10^{3}$8) (Middelberg et al., 2010, Herzog et al., 2016). Many IFRSs display compact morphologies (unresolved at arcsecond scale), but a subset show extended double-lobe (FR II-like) structures (Collier et al., 2013, Singh et al., 2017). VLBI observations reveal that most IFRSs contain AGN cores with high brightness temperatures ($10^{3}$9 K), confirming their non-thermal, compact nature (Herzog et al., 2015).

Infrared and Optical: The defining characteristic remains the non-detection or extreme faintness at 3.6/3.4 μm, with most sources lying just above the survey limits ($z\sim2-4$0–$z\sim2-4$1Jy in the deepest fields) (Norris et al., 2011). At higher fluxes, only a minority are detected in WISE W1 (3.4 μm) at $z\sim2-4$2–$z\sim2-4$3Jy (Collier et al., 2013). Optical counterparts are extremely rare or, if present, typically have $z\sim2-4$4 (Singh et al., 2017). Far-IR and submillimetre limits from Herschel and Spitzer preclude substantial cold-dust emission in most IFRSs, with stacking analysis yielding non-detections at $z\sim2-4$5 mJy (Herzog et al., 2015).

Polarization, X-ray, and SEDs: IFRSs show significant radio polarization (median $z\sim2-4$6\%, up to $z\sim2-4$7\%), comparable to lobe-dominated AGN (Collier et al., 2013). X-ray counterparts are extremely rare; when detected, they confirm a Type 1 AGN character. SED fitting, including Bayesian analysis, conclusively demonstrates that an AGN component is required to explain all data; starburst or non-AGN dust SEDs cannot fit the extreme radio-IR properties (Zhang et al., 2024, Huynh et al., 2010). In the FIR, the best-fit SEDs imply total $z\sim2-4$8 in the $z\sim2-4$9–$S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$0 $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$1 range, consistent with the ULIRG regime but generally below classical HzRGs (Herzog et al., 2015).

3. Redshift Distribution and Host Galaxy Context

Spectroscopically confirmed IFRSs have redshifts primarily in the range $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$2, with a median $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$3–$S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$4 (Herzog et al., 2013, Orenstein et al., 2018, Collier et al., 2013, Singh et al., 2017). No confirmed IFRS has $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$5. Lower IR flux densities correspond to higher redshifts due to the empirical $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$6–$S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$7 anti-correlation, fitted as $S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$8Jy (Orenstein et al., 2018). This mapping enables the use of IR flux thresholds to identify higher redshift radio AGN candidates.

Host galaxies are massive ($S_{1.4\,\mathrm{GHz}} \gtrsim 0.1$9–$\sim5\,\mu$0), often with significant dust, and are frequently undetected at optical and $\sim5\,\mu$1-band depths (Huynh et al., 2010, Zhang et al., 2024, Singh et al., 2017). SED results show both Type 1 QSO-like and AGN–starburst composite SEDs in the IFRS population. No direct evidence supports a dominant starburst or non-AGN power source.

4. Physical Nature: AGN, Evolutionary State, and Radio Properties

AGN Signature: All radio, SED, and VLBI findings affirm that IFRSs are predominantly AGN-powered, with compact cores and steep-spectrum synchrotron emission (Herzog et al., 2015, Middelberg et al., 2010). No bona fide IFRS has been shown to be a pulsar, extended starburst, or Galactic object (Cameron et al., 2011).

Spectral Properties: Steep spectra ($\sim5\,\mu$2) are ubiquitous; a minority of IFRSs are ultra-steep (USS; $\sim5\,\mu$3), a known tracer of high-redshift radio AGN (Herzog et al., 2016, Middelberg et al., 2010). Some IFRSs are classified as Gigahertz-Peaked Spectrum (GPS) or Compact Steep Spectrum (CSS) sources, corresponding to young, compact AGN at stages before full-scale FR I/II morphology develops (Collier et al., 2013, Herzog et al., 2015, Herzog et al., 2016).

Morphological Diversity: IFRSs exhibit both compact (unresolved $\sim5\,\mu$4 arcsec, $\sim5\,\mu$5 5–10 kpc physical size at $\sim5\,\mu$6) and classical double-lobe morphology. VLBI detection rates of $\sim5\,\mu$761\% [57/35 detected] are significantly higher than in generic radio AGN samples (Herzog et al., 2015). There is evidence for a positive correlation between core compactness and redshift in the IFRS population, consistent with an evolutionary sequence from compact, young sources toward more extended morphologies at lower redshift or higher radio luminosity (the GPS$\sim5\,\mu$8CSS$\sim5\,\mu$9FR I/II pathway).

Infrared Emission and Star Formation: Stacking analyses and FIR limits constrain star formation rates to $\geq$0/yr for the majority of IFRSs, with SEDs generally dominated by AGN torus-heated dust (Zhang et al., 2024, Herzog et al., 2015). For examples with significant FIR detections, AGN and star-forming dust contributions are comparable, but most sources are AGN-dominated. There is no significant observed correlation between AGN luminosity and SFR within the IFRS sample.

5. Role in Galaxy Evolution, Cosmology, and Surveys

IFRSs significantly extend the census of high-redshift, radio-loud AGN and offer a window onto SMBH and massive galaxy assembly at $\geq$1 (Herzog et al., 2013, Zinn et al., 2011, Orenstein et al., 2018). Their sky density ($\geq$230 deg$\geq$3) implies a much larger high-$\geq$4 AGN population than inferred from classical, optically-selected HzRG samples (surface density $\geq$50.001 deg$\geq$6). This amplifies challenges for galaxy formation models, particularly in accounting for the rapid build-up of SMBHs shortly after the Big Bang.

IFRSs are also crucial for studies of the radio AGN luminosity function, feedback processes, and the cosmic X-ray background (CXB). The inferred comoving SMBH mass density associated with IFRSs is $\geq$7–$\geq$8, sufficient to explain a significant fraction of the unresolved soft and hard CXB (Zinn et al., 2011).

A summary of key physical and survey properties:

Quantity	Typical Value	Reference
$\geq$9	$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$0 mJy	(Collier et al., 2013, Filipović et al., 2021)
$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$1	$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$2–$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$3Jy	(Norris et al., 2011, Collier et al., 2013)
$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$4	$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$5 (typ. $$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$6)	(Norris et al., 2011, Zinn et al., 2011)
Median $$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$7	$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$8–$$\frac{S_{1.4\,\mathrm{GHz}}}{S_{3.6\,\mu\mathrm{m}}} > 500$$9	(Orenstein et al., 2018, Herzog et al., 2013)
Radio spectral index	$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$0 to $S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$1	(Herzog et al., 2016, Middelberg et al., 2010)
VLBI core detection rate	$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$2	(Herzog et al., 2015)
Host stellar mass	$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$3	(Zhang et al., 2024, Huynh et al., 2010)
FIR detection (Herschel)	not detected, $S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$4	(Herzog et al., 2015)
SFR upper limit	$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$5/yr	(Herzog et al., 2015, Zhang et al., 2024)

6. Population Diversity and Selection Function

The IFRS class is heterogeneous. The observed SED dichotomy—half displaying Type 1 QSO-like UV/optical SEDs, the other half resembling AGN–starburst composite IR SEDs—suggests at least two sub-populations or evolutionary phases within IFRSs. The most extreme, IR-faintest IFRSs are likely at the highest redshifts ($S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$6), perhaps representing progenitors of massive radio galaxies and the earliest phases of SMBH growth (Maini et al., 2016, Herzog et al., 2013). The infrared selection function is inherently redshift-dependent, such that fainter $S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$7 selects higher-$S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$8 AGNs; using empirical $S_{3.6\,\mu\mathrm{m}} < 30\,\mu\mathrm{Jy}$9–$R$0 fits enables extension to $R$1 in future radio–IR surveys (Orenstein et al., 2018).

Pragmatic recommendations for selecting high-$R$2 AGN, based on both modeling and empirical results (Maini et al., 2016, Orenstein et al., 2018), are:

• $R$3–$R$4 mJy with $R$5 and $R$6 selects $R$7 RL AGN.
• The canonical [Zinn et al.] $R$8 cut efficiently yields $R$9–$\sim 500$0 AGNs.

7. Future Directions and Open Questions

Ongoing and upcoming surveys—EMU/ASKAP, VLASS, LoTSS, MIGHTEE—will expand IFRS samples to fainter flux limits and wider areas (Herzog et al., 2016, Collier et al., 2013). Open research directions include:

• Securing spectroscopic redshifts for IR-faintest IFRSs; photometric redshifts via JWST or ALMA millimeter spectroscopy for $\sim 500$1 candidates.
• Detailed host-galaxy characterization: stellar populations, dust properties, environments.
• High-resolution (VLBI) imaging to map radio core-jet structure and clarify the fraction of CSS/GPS-like subtypes.
• Refining evolutionary pathways: quantifying connections between IFRSs, young RLAGN, and extended FR I/II radio galaxies.
• Assessing the impact of IFRSs on CXB modeling and SMBH mass function evolution at cosmic dawn.

---

Infrared-Faint Radio Sources thus constitute a quantitatively distinct, physically meaningful, and cosmologically valuable population of high-redshift, radio-loud AGN, representing both an efficient means to trace SMBH growth in the early universe and an astrophysical laboratory for the study of AGN formation, feedback, and obscured star formation (Norris et al., 2011, Zhang et al., 2024, Orenstein et al., 2018, Herzog et al., 2013, Herzog et al., 2015).