Rogue Worlds Drift Through the Milky Way in the Billions

The Empty Space Between Stars Is Full of Something

A single lensed star brightened for less than a day in May 2024. An unseen object passed between it and Earth, bending and magnifying its light just enough for telescopes in Chile, South Africa and Australia to catch the signal. Gaia, orbiting far from both, caught it too, with a measurable time offset that let astronomers pin down the object’s mass: roughly 0.22 Jupiter masses, or about 70 Earth masses. Slightly less massive than Saturn. No host star was detected.

This one event, reported by Subo Dong and 41 colleagues in Science in January 2026, is not a census. It is a single data point proving that at least one planetary-mass object in the Milky Way’s halo can exist without a star. But it arrives on the heels of a broader statistical claim that changes how we should think about the galaxy: free-floating or very wide-orbit planets may outnumber stars by roughly 20 to one.

That figure comes from a different paper, led by Takahiro Sumi and published in The Astronomical Journal in 2023. It analysed nine years of observations from the MOA-II survey toward the Galactic bulge and found six short-duration microlensing events consistent with planetary-mass lenses. When the team fitted a power-law mass function across objects from roughly a third of Earth’s mass up to 21 Jupiter masses, the result came out to 21 free-floating or very wide-orbit objects per star, with a one-sigma range stretching from about eight to 44.

Twenty to one is the rounded centre of a broad distribution, not an exact tally. And it is dominated by the low-mass end, where the power law implies many more tiny worlds than massive ones. That detail matters, because another model published in 2025 predicts the distribution turns over well below that threshold. Gavin Coleman and William DeRocco, writing in Monthly Notices of the Royal Astronomical Society, simulated planet formation and ejection across single-star, close-binary and wide-binary systems and arrived at a prediction of just over one free-floating planet per star over essentially the same mass range. Their model agrees with the observed slope above about eight Earth masses but then predicts a flattening, leaving far fewer small rogue worlds than the 2023 analysis implies.

The two studies are not reconciled yet. What they agree on is that something is out there in large numbers, and that whatever it is reshapes our picture of how planetary systems assemble and scatter.

How One Event Became a Mass, Not Just a Light Curve

Microlensing is clever but degenerate. A short brightening event tells you the Einstein crossing time, which scales with the square root of the lens mass—but duration alone cannot fix the mass. Distance, relative velocity and actual mass can trade off against each other to produce the same light curve.

The May 2024 event had two things working in its favour. First, it showed a rounded peak caused by a finite-source effect: the lens crossed the apparent disc of a background red giant rather than a point source, and from that the team derived an angular Einstein radius of 18.6 microarcseconds. That removed one unknown.

Second, and crucially, Gaia happened to scan the same patch of sky six times over a 16-hour window. The spacecraft orbits near the Sun-Earth L2 point and cannot be redirected on demand, so this alignment was not planned. Still, the data contained the offset needed for a parallax measurement: the event peaked about 1.9 hours later in Gaia’s data than it did in the ground-based light curves, because the two observers viewed the alignment from separated positions. Fitting that offset gave a parallax detection at roughly seven sigma.

Combining parallax with the angular Einstein radius broke the mass-distance degeneracy and yielded a direct gravitational mass estimate: 0.219 Jupiter masses, plus or minus 0.075 and minus 0.046. The inferred distance was 3.05 kiloparsecs, or about 10,000 light-years, with an uncertainty of roughly plus or minus half a kiloparsec.

No direct image was taken. The object cannot be seen in these data at all. The mass is gravitational, inferred from how the lens warped spacetime around a background star.

Ejected, Not Born Alone

The Science paper describes the object as either gravitationally unbound or on a very wide orbit whose host star left no detectable lensing signature. Calling it a rogue planet is shorthand, and a useful one, but not every such candidate can be fully disentangled from a distant star in a single brief event. Future high-resolution imaging may eventually resolve the foreground lens from the background source as they drift farther apart on the sky, and could reveal a faint companion if one exists.

Dong’s team argues that the object’s mass and its place in the wider microlensing distribution are most consistent with formation inside a protoplanetary disc followed by dynamical ejection, rather than collapse like a brown dwarf. That is an inference, not an observation. The mass function implied by the 2023 population analysis leans toward the same story, but Coleman and DeRocco’s 2025 simulations show that the full picture depends heavily on how binary systems contribute to ejection rates, and on whether the mass function really continues rising steeply down to Earth-mass scales.

Saturn-mass does not mean Saturn-sized. Microlensing measures gravitational influence, not radius, atmosphere, rings or surface temperature. What we have is a weight, a distance and a trajectory through the Galaxy, not a world we can look at.

What Changes When the Census Shifts

If the 20-to-one estimate holds up—and it does not yet—the implications are not just numerical. They are structural.

Planetary formation theory already predicts ejection. Simulations of the early solar system show that Jupiter and Saturn likely scattered smaller bodies outward, and that several planetary embryos were probably thrown free during the chaos of assembly. A 20-to-one ratio simply means that ejection is not a rare side effect but a dominant outcome: most planets do not stay where they formed. That revises the baseline for how we think about the Milky Way’s inventory of worlds. Stars are not the only repositories of planetary mass; the dark interstellar medium between them may carry far more of it.

It also changes the calculus for indirect searches for extraterrestrial life. Free-floating planets are not automatically lifeless. A world ejected from its system could retain internal heat from radiogenic decay and tidal flexing long after it lost its star. Subsurface oceans beneath thick ice shells are plausible on bodies this size, and the energy required to keep liquid water alive does not strictly depend on a nearby sun. Ocean worlds like Europa and Enceladus already demonstrate that biology-friendly environments can exist far from any star. Multiply that possibility by tens of billions of rogue objects in the Milky Way and the volume of potentially habitable real estate grows in a way that star-centric models underestimate.

At the same time, detecting life on such objects would be astronomically harder. A free-floating planet emits almost no reflected light. Any biosignature would have to be inferred from thermal emission, atmospheric spectra or, eventually, from technology-assisted observation. Microlensing itself can provide spectroscopy in principle, but the brief, rare alignments make it an inefficient probe of atmosphere composition. We are not close to characterising the atmospheres of these objects at scale.

The Next Decade of Answers

The Nancy Grace Roman Space Telescope’s microlensing programme is designed to address exactly these uncertainties. With a much larger sample and sensitivity extending to lower-mass lenses, Roman will turn a handful of well-measured events into a distribution. The critical test will be the shape of the mass function below Earth mass: does it keep rising, as the 2023 power-law fit suggests, or does it turn over, as the Coleman-DeRocco simulations predict? Either answer would constrain how common ejection was and what fraction of planetary mass is locked in free-floating objects versus bound systems.

Individual follow-up will also improve. For events like the one Dong’s team studied, where finite-source effects and parallax both contributed clean measurements, later imaging with the James Webb Space Telescope or future missions could finally resolve the lens from the source and search for any companion star. A resolved companion would reclassify the object from rogue to wide-orbit planet. No companion would strengthen the free-floating case.

The broader lesson is that the Milky Way is more crowded with planets than we used to think, and far less orderly. Most stars probably once held systems that were destabilised by mutual gravity, by passing stars or by the density of the galactic disk itself. The planets that survived are the ones that remained bound. The ones that did not are still here, drifting, cooling, carrying whatever geology and chemistry they retained at ejection into the dark between stars.

A single Saturn-mass object does not prove the 20-to-one claim. But it proves the category exists. And the category, if the statistics hold, is enormous.

What happens next depends on whether Roman delivers enough events to distinguish between a steeply rising mass function and one that flattens in the terrestrial regime. If the former is right, rogue worlds dominate the galaxy’s planetary inventory. If the latter is right, most free-floating objects are sub-Earth in mass and the bulk of planetary building material stays bound to stars. Either outcome changes how we interpret the Milky Way’s history, and either one makes the dark space between stars more interesting than it looked a decade ago.