Which exoplanets could the SKA hear? A radio forecast with important caveats

Jupiter’s auroras produce radio bursts that ground-based telescopes can detect. Finding comparable emissions from an exoplanet could offer a way to investigate its magnetic field and rotation. The challenge is deciding where to look—and whether a predicted signal would actually reach Earth.

A study by Mahdiyar Mousavi-Sadr, Fatemeh S. Tabatabaei, Alexander Wolszczan and Ghassem Gozaliasl tackles that selection problem. Its revised arXiv preprint, listed as submitted to Publications of the Astronomical Society of Australia, forecasts radio emission rather than reporting new detections. The paper states that no exoplanetary radio detection has yet been confirmed.

From stellar winds to radio forecasts

The researchers assembled 1,027 confirmed planets orbiting F-, G-, K- and M-type main-sequence stars older than 700 million years. That age restriction matches the stellar-wind relations used in their calculations. They estimated missing radii for 109 planets using a mass–radius relationship and calculated missing orbital separations for 59 using Kepler’s third law.

Their main tool is the radiometric Bode’s law, a semi-empirical relationship calibrated on radio-emitting solar-system planets. In this model, a stellar wind supplies energy to a planetary magnetosphere, some of which becomes auroral radio emission. The calculations estimate both the flux density reaching Earth and a characteristic maximum emission frequency related to the planet’s magnetic field.

Rather than assuming a constant wind speed, the team used Parker’s isothermal wind model to account for acceleration with distance from the star. This matters for close-orbiting planets, where an assumed fully accelerated wind could overestimate the available radio power. The model also assumes tidal locking for planets orbiting within 0.1 astronomical units.

Fifteen nominal targets

Most of the sample fails the first observational hurdle: 794 planets have predicted characteristic frequencies below about 10 megahertz, where Earth’s ionosphere blocks ground-based observations. An additional 87 fall outside the study’s adopted practical SKA sky coverage, leaving 146.

Applying the planned telescope frequency bands narrows that group to 53 planets for SKA-Low and 12 for SKA-Mid. The researchers then compare predicted brightness with five-times-noise detection thresholds, assuming one-hour observations and a bandwidth equal to approximately 30% of the observing frequency.

Under those assumptions, 13 planets qualify for the full-design AA4 configuration of SKA-Low and two for SKA-Mid. Nine of those 15 also qualify for the intermediate AA* configuration: seven Low targets and both Mid targets.

WASP-14 Ab leads the SKA-Low list, with a predicted flux density of about 40.1 millijanskys and a characteristic frequency of 57.5 megahertz. HD 141937 b is highlighted for SKA-Mid, with 0.084 millijanskys at 733.1 megahertz. These are calculated values, not measured signals.

Why the shortlist is provisional

Eight of the 15 candidates lack directly measured radii. Propagating the bounds of the adopted mass–radius relationship produces an average spread of approximately 0.94 dex—nearly an order of magnitude—in predicted flux and frequency across the 109 radius-estimated planets.

For the eight shortlisted planets, none satisfies every detection requirement across the full propagated uncertainty range. The authors therefore classify them as marginal under this stricter test. That group includes HD 141937 b, making its status as a leading Mid target explicitly conditional.

A different problem affects TOI-4515 b. It lies in a mass-and-orbital-distance region where an expanded, ionised atmosphere may suppress auroral radio emission. Its nominal brightness therefore does not necessarily translate into an observable signal.

The underlying physics is approximate too. The model estimates magnetic moments from mass and rotation rather than modelling planetary interiors and dynamos directly. It cannot reliably distinguish an active magnetic dynamo from an inactive one in planets with similar bulk properties.

Machine learning explains the model, not observations

Permutation importance and SHAP analyses identify planetary radius, orbital separation, stellar age and Earth–star distance as the strongest inputs for reproducing the modelled fluxes. A random forest using those four inputs achieves a test-set coefficient of determination of 0.880, close to 0.883 with eight inputs.

But its training targets are synthetic outputs from the same physical model, not observed exoplanet radio signals. It is an emulator, not independent confirmation. Planetary mass also remains necessary for assessing characteristic frequency and hence detectability.

Finally, auroral emission is beamed and variable. A bright source may point away from Earth or remain quiet during an observation. The paper argues for repeated or phase-targeted observations and searches in circular polarisation. The shortlist provides targets for testing the predictions—not evidence that those planets have already been heard.