Somewhere in the Milky Way tonight, drifting through the darkness between the stars, there are roughly 20 planets for every sun that burns — trillions of worlds, most thought to be ejected from the solar systems where they were born and now wandering the galaxy, according to a nine-year microlensing survey

A nine-year microlensing survey has produced the most complete count yet of the small, dark, planet-mass objects that drift through the Milky Way without a star. The estimate lands at roughly 20 for every star in the galaxy, most of them low in mass, some down around the mass of Earth. Set against the several hundred billion stars in the Milky Way, that puts the total in the trillions.

The number comes from the MOA collaboration, led by Takahiro Sumi of Osaka University, in a paper published in The Astronomical Journal in 2023. It draws on data collected between 2006 and 2014 by the MOA-II survey at Mount John University Observatory in New Zealand, which watched the crowded star fields toward the centre of the galaxy for the faint, brief brightenings that betray an unseen mass.

How the count is made

Microlensing is an indirect method, and an unforgiving one. When a foreground object passes almost exactly in front of a more distant star, its gravity bends the background starlight and magnifies it for a short time. The heavier the lensing object, the longer the event lasts. A star produces one that runs for weeks. A planet-mass object produces one that can be over in a day or less.

Those very short events are the whole story here. Out of thousands of microlensing candidates in the MOA-II dataset, the free-floating planet estimate rests on just six with crossing times under half a day. The shortest, described in a companion paper led by Naoki Koshimoto as only the second Earth-mass candidate found to date, lasted about 0.057 days, a little under an hour and a half. From that small set, and a careful accounting of how many such events the survey would have missed, the team worked out how common the underlying objects must be to produce what was actually seen.

The result is a rate of about 21 planet-mass objects per star, with a stated uncertainty running from roughly 8 to 44. That is a wide range, and the paper does not hide it. The central figure is striking, but the error bars are part of the finding, not a footnote to it.

Free-floating, or just very far out

The paper is precise about what it is counting: free-floating planets or planets in very wide orbits. Microlensing sees the lens, not its surroundings. When no host star shows up in the signal, it can mean the object is truly unbound, ejected from the system where it formed and moving through the galaxy on its own. It can also mean a planet still loosely tied to a star too far away to register in an event this brief.

Most of these objects are thought to be genuine wanderers, thrown out by gravitational tussles during the early, unstable years of a planetary system. That is the leading interpretation, and it fits the numbers well. Even so, the distinction is easy to lose in the retelling.

A correction to an older claim

The more interesting part of this result, in our reading, is what it revised. In 2011 the same survey reported evidence for a large group of free-floating planets around the mass of Jupiter, outnumbering stars by roughly two to one. That claim circulated widely.

The 2023 analysis, with three more years of data and a more careful treatment of detection efficiency, does not support it. The heavy Jupiter-mass excess largely disappears. What remains is a far greater number of much lighter objects, weighted toward the low end of the mass scale. The headline count went up, but the picture underneath it changed: fewer giants, many more small bodies.

This is one survey, not a settled census, and the low-mass end depends on very few detections. The measurement is a real one. It is not, on its own, the final tally of how many of these objects the galaxy holds.

What Roman is expected to add

The clearest test is already funded and under construction. NASA’s Nancy Grace Roman Space Telescope, due to launch by May 2027, will run a microlensing survey from space, above the blur of Earth’s atmosphere and with a much wider field of view. In coverage of the MOA work, NASA estimated that Roman could detect about 400 Earth-mass free-floating planets, and that observing from space would make it sensitive to lower-mass objects than ground-based surveys can reach.

That would move the field from an inference built on a handful of fleeting events to a sample large enough to map the mass distribution properly. If the low-mass objects are really as common as the MOA estimate implies, Roman should see them plainly. If they are not, the same survey will show that too.

For now the estimate stands at about 20 per star, built from nine years of observation and a few of the shortest events MOA recorded. Whether it holds will depend on Roman, and on ground-based partners such as Japan’s PRIME telescope in South Africa, which is beginning the first wide-area microlensing survey in near-infrared light. Once those instruments have looked at the same fields, the count will either firm up or come down.