Deep Radio Survey Reveals Hidden Galaxies and Black Holes

The Sky That Wouldn’t Stop Talking

For decades, when astronomers pointed their radio telescopes at the sky, they were listening to the loudest voices. Bright quasars. Violent radio galaxies. The cosmic equivalent of a stadium concert. The quiet murmurs, the background conversations, the galaxies just going about their business of making stars, those were mostly invisible.

That has changed.

A team led by Philip Best from the University of Edinburgh has released the deepest, widest radio survey ever conducted. Using the Low Frequency Array (LOFAR), a network of radio telescopes scattered across Europe, they mapped a patch of sky roughly the size of 200 full moons. And what they found upends the old story of what radio astronomy sees when it looks at the universe.

The old story was simple: radio sources are mostly black holes. Active galactic nuclei (AGN) blasting out jets of plasma, screaming across the spectrum. The new story, based on the LOFAR Deep Fields data, is that the sky is full of ordinary galaxies. Quiet ones. Star-forming ones. The kind of places where planets and life might exist.

More than two-thirds of the radio sources in this survey are star-forming galaxies, ranging from normal galaxies nearby to highly starbursting systems at redshifts greater than 4 (Best et al., 2023). The radio sky, it turns out, is not a black hole zoo. It is a nursery.

How to See the Invisible Majority

The Problem with Radio Surveys

Radio waves from space are hard to catch. They are long, low energy, easily drowned out by the radio chatter of human civilization. Early radio surveys were like trying to map a forest at night with a single flashlight: you saw the biggest trees, the ones right in front of you, and missed everything else.

The LOFAR Deep Fields survey changed the game by being patient. Really patient. The team observed the same patches of sky for hundreds of hours, stacking the data until faint signals emerged from the noise. The result is a catalog of roughly 80,000 radio sources, each one corresponding to a galaxy somewhere in the universe.

But finding the galaxies was only the first step. The hard part was figuring out what kind of galaxy each one was.

The Four Flavors of Radio Galaxy

The team classified every source into one of four categories (Best et al., 2023):

• ▸Star-forming galaxies: Normal galaxies making stars, including starbursts that are forming stars at extreme rates
• ▸Radio-quiet AGN: Galaxies with a black hole that is active but not producing powerful radio jets
• ▸Radio-loud high-excitation AGN: Black holes that are actively feeding and producing strong radio emission
• ▸Radio-loud low-excitation AGN: Black holes that are radio-loud but not currently feeding much

To do this, they used four different computer codes to model each galaxy's light across the spectrum, from ultraviolet to far-infrared. Two of the codes included models of AGN activity. By comparing the results, they could tell whether a galaxy's radio emission came from star formation, from a black hole, or from some combination of both.

Ninety-five percent of the sources could be reliably classified (Best et al., 2023). That is an astonishing success rate for a field where ambiguous classifications are the norm.

The Surprising Demographics of the Radio Universe

The 1 Millijansky Threshold

If you look at bright radio sources, above about 1 millijansky at 150 MHz, the story is what astronomers expected: AGN dominate. Black holes are the loudest voices in the room.

Below that threshold, something shifts. Suddenly, star-forming galaxies take over. At flux densities around 100 microjanskys, star-forming galaxies account for 90 percent of all radio sources (Best et al., 2023).

This is the radio equivalent of realizing that the bright streetlights you've been studying are actually rare, and most of the city is lit by ordinary windows.

The Quiet Black Holes

Radio-quiet AGN are the most surprising population in the survey. They make up about 10 percent of the total, and they are hard to find. These are galaxies where the central black hole is active, producing X-rays and optical emission, but not producing strong radio jets. They are black holes whispering, not shouting.

The fact that radio-quiet AGN exist in large numbers tells us something important about how black holes grow. Many black holes go through phases where they are feeding but not producing jets. Those phases might be common, even if they are hard to detect.

The authors found that radio-quiet AGN are often missed by traditional radio surveys because their radio emission is weak. Only by combining radio data with multi-wavelength observations can you identify them (Best et al., 2023).

Why This Changes Everything

The Star Formation History of the Universe

One of the biggest questions in astronomy is: how did galaxies build their stars over cosmic time? We know that star formation peaked about 10 billion years ago and has been declining ever since. But measuring star formation rates accurately across cosmic time is hard.

Radio emission tracks star formation in a way that optical light cannot. Young, massive stars produce supernovae, which accelerate electrons, which produce radio waves. The radio signal is unaffected by dust, which blocks optical light. So radio surveys give a clean view of star formation, even in the dustiest galaxies.

The LOFAR Deep Fields survey has found star-forming galaxies out to redshifts beyond 4, meaning we are seeing them as they were when the universe was less than 2 billion years old (Best et al., 2023). These are the building blocks of modern galaxies, caught in the act of construction.

The Simulation Problem

Astronomers use computer simulations to predict what the radio sky should look like. Two major simulations, SKADS and T-RECS, have been used to design future radio telescopes and interpret existing data.

The LOFAR data shows that both simulations are wrong in important ways. The authors found that the simulations overpredict the number of radio-loud AGN at faint flux densities and underpredict the number of star-forming galaxies (Best et al., 2023).

This matters because we are building the Square Kilometre Array, the largest radio telescope ever conceived. If the simulations guiding its design are off, we might miss the most interesting science.

The authors suggest specific improvements to the simulations, including better models of how star formation produces radio emission and more accurate treatments of AGN evolution (Best et al., 2023).

How They Did It: The Technical Core

The Data

LOFAR is not a single telescope. It is a network of thousands of antennas spread across the Netherlands and other European countries. By combining signals from all these antennas, LOFAR can achieve the resolution of a telescope hundreds of kilometers across.

The Deep Fields survey targeted three areas of sky that had already been studied extensively by other telescopes, including the Hubble Space Telescope and the Spitzer Space Telescope. This meant the team had deep optical and infrared data to match against their radio sources.

The Analysis

For each of the 80,000 radio sources, the team measured the galaxy's light across 20 or more wavelengths. They then fitted this data with four different computer models, each one making different assumptions about how galaxies work.

Two of the models, CIGALE and MAGPHYS, assume that all the light comes from stars and star formation. Two others, AGNFitter and X-CIGALE, include the possibility that some light comes from an actively feeding black hole.

By comparing the results, the team could identify galaxies where an AGN model fit better than a pure star formation model. They could also identify galaxies where the radio emission was stronger than expected from star formation alone, flagging them as AGN candidates.

The Classification

The final classification into four types used a combination of methods. Galaxies with strong radio emission relative to their star formation rate were classified as radio-loud AGN. Among those, galaxies with high-excitation emission lines were classified as high-excitation AGN, while those without were classified as low-excitation AGN.

Galaxies with radio emission consistent with star formation were classified as star-forming galaxies. Those with evidence of an AGN from optical or X-ray data but weak radio emission were classified as radio-quiet AGN.

The result is a clean, physically meaningful classification for 95 percent of the sources (Best et al., 2023).

What the Research Does Not Prove

This is an observational survey, not an experiment. The team cannot prove that star formation causes the radio emission in every case, or that the AGN classifications are correct for every source. Some galaxies might be misclassified.

The survey covers only three patches of sky. While these patches were chosen to be representative, the universe is a big place. There could be large-scale variations in galaxy populations that this survey misses.

The classification relies on models, and models have assumptions. The four SED fitting codes sometimes disagree, especially for galaxies with complex histories. The team used a consensus approach to resolve disagreements, but consensus is not proof.

Finally, the survey sees galaxies at different distances. A galaxy at redshift 4 is seen as it was 12 billion years ago. A galaxy at redshift 0.1 is seen as it was 1 billion years ago. Comparing these populations requires careful correction for cosmic evolution, and those corrections depend on models that are not fully tested.

These are not fatal flaws. They are the normal uncertainties of frontier science. Every claim in the paper is made with appropriate caution, and the authors are explicit about their methods and assumptions.

What This Actually Means

• ▸The next generation of radio telescopes, including the Square Kilometre Array, should prioritize sensitivity over raw collecting area. The most interesting science at faint flux densities is not black holes but star-forming galaxies. Build telescopes that can see the quiet majority.

• ▸Galaxy evolution models need to account for radio-quiet AGN. These black holes are feeding and growing, even if they are not producing jets. Ignoring them means underestimating the total energy input from black holes into galaxies.

• ▸Star formation rates derived from radio data are now reliable out to high redshift. The LOFAR Deep Fields catalog provides a benchmark for calibrating other star formation indicators. If your optical star formation rate disagrees with the radio, trust the radio.

• ▸The SKADS and T-RECS simulations need revision. Anyone using these simulations to design surveys or interpret data should be aware of the discrepancies found by Best et al. (2023). The updated simulations should include better AGN models and more realistic star formation prescriptions.

• ▸The universe is more ordinary than we thought. The cosmic radio sky is not dominated by exotic black hole engines. It is dominated by galaxies making stars, just like our own Milky Way. The spectacular is real, but it is rare. The quiet majority is what we should be studying if we want to understand how galaxies live and die.