For decades, astronomers have often treated the birth of stars as if the Milky Way were working from a remarkably consistent recipe: given a cloud of gas and dust, nature would produce roughly the same distribution of stellar masses again and again. Now, a new study using observations of hundreds of Milky Way star clusters offers direct evidence that the recipe is not universal. Instead, the mass distribution of newborn stars appears to have changed across the galaxy’s history, with older star clusters showing a different characteristic mass than younger ones. The result comes from a careful analysis of stellar populations observed by the European Space Agency’s Gaia mission, and it points toward a picture in which the conditions inside the clouds where stars form help determine what kinds of stars are born.
When astronomers look at a galaxy, they see a dazzling mixture of stars. But that light can hide an important detail: most of the stellar mass is not concentrated in the rare, brilliant stars that dominate what we see.
To understand the contents of a galaxy, astronomers therefore need to know how stars were distributed by mass when they formed. That distribution is called the stellar initial mass function, or IMF.
The IMF is essentially a birth record for stars. It tells astronomers how many stars form at different masses in a newly created stellar population.
For a long time, astronomers have commonly assumed that this distribution is universal. The same basic IMF has been used to interpret stellar populations under very different conditions, both inside the Milky Way and in distant galaxies.
That assumption has been useful. It has also been difficult to test.
The problem is that most galaxies are too distant for astronomers to measure their individual stars directly, especially the numerous low-mass stars that are crucial for determining the IMF. Even within the Milky Way, measuring the IMF is challenging because a cluster observed today is not necessarily a pristine snapshot of the stars that were born there.
Stars die. Clusters lose stars. And their environments change.
The new study takes advantage of a particularly useful feature of open clusters: astronomers can observe their individual stars and reconstruct their present-day stellar mass distributions.
The researchers then looked for something that could separate the original stellar birth pattern from everything that happened afterward.
They found it in the break mass.
The clue hidden in a cluster’s stellar population
A stellar mass function does not follow one simple slope across all stellar masses. In the Kroupa IMF used by the researchers, the distribution changes behavior around a characteristic mass.
That transition, or break, becomes especially useful because different physical processes affect it in different ways.
Imagine a cluster as a stellar city that has been changing for millions or billions of years.
The most massive stars have short lifetimes, so some disappear as they evolve. Meanwhile, the cluster’s gravitational environment can gradually strip away low-mass stars, particularly through tidal effects. These processes alter the cluster’s present-day population.
But according to the models used in the study, those processes change the slopes of the mass function without moving its break mass.
That distinction is crucial.
If the break moves from one cluster to another, the researchers argue, ordinary dynamical evolution cannot explain the shift. A difference in the break mass therefore provides a way to look back toward the IMF at the time the cluster formed.
This is also why the researchers specifically used a Kroupa-like mathematical description of the IMF. They found that dynamical depletion can modify its power-law slopes while leaving the break mass essentially unchanged. A Chabrier-like or log-normal description does not provide the same clean separation: dynamical evolution can shift the inferred characteristic mass and create a degeneracy with genuine IMF variation.
That gave the researchers a testable question:
Do different open clusters actually have different break masses?
The Gaia data say yes.
Gaia turns the Milky Way into a laboratory
The study began with a Gaia DR3 catalog containing 7,167 detected clusters, including 3,530 highly reliable open clusters.
But having thousands of clusters was not enough.
The researchers needed clusters in which both the low-mass stars below the break and the rarer high-mass stars above it could be measured well enough to determine the shape of the mass distribution.
After applying a series of quality requirements, the final analysis contained 110 open clusters.
The restrictions were demanding. The researchers required reliable cluster identification, sufficient mass bins on both sides of the break, strong statistical significance, high-quality color-magnitude diagrams, relatively low extinction, suitable mass-function slopes, accurately measured break masses and reliable ages.
The result was a much smaller sample than the full Gaia cluster catalog, but one in which the researchers could make a much more precise measurement.
Only 417 of the 3,530 high-quality clusters had enough mass completeness to detect stars below the break in the first place. And only a subset of those also contained enough of the relatively rare high-mass stars needed for the analysis.
The final 110 clusters therefore represented a carefully selected population rather than a random sampling of every known cluster.
The break moves with cluster age
The striking pattern emerged when the researchers compared the break mass with the ages of the clusters.
The break mass was strongly correlated with age.
In the raw sample, the Spearman correlation coefficient was 0.391, with a reported p-value of 2.4 × 10⁻⁵. After the researchers modeled a distance-dependent selection effect, the correlation became stronger, reaching 0.449, with a p-value of 8.7 × 10⁻⁷.
In simple terms, older clusters tended to have higher break masses.
That is the central observation.
And it is important what did not change in the same way.
The researchers found no statistically significant correlation between cluster age and either the high-mass slope or the intermediate-mass slope in their main sample. The high-mass slope had a Spearman correlation of 0.055 with a p-value of 0.57, while the intermediate-mass slope had a correlation of −0.153 with a p-value of 0.06.
The strongest age-dependent signal was therefore concentrated in the break mass rather than in the slopes.
That pattern matters because it matches what the researchers expected if the IMF itself had changed.
Why the researchers don’t think aging clusters caused the pattern
At first glance, an older cluster having a different stellar population might seem unsurprising.
After all, an old cluster has had much more time to lose stars and for its massive stars to evolve away.
But those effects do not provide the explanation the researchers are looking for.
Their modeling indicates that dynamical depletion is approximately scale-free: it changes the slopes of the mass function but does not move the Kroupa break mass.
The same distinction appears in their analysis of individual clusters.
For NGC 6067, for example, the raw stellar counts already showed a break at 1.23 ± 0.02 solar masses. After corrections for Gaia selection effects and unresolved binaries, the measured break was 1.28 ± 0.14 solar masses.
The corrections changed the slopes substantially more than they changed the location of the break.
The researchers also tested what happened if they deliberately assigned the cluster the wrong age. Using an age of 4 billion years instead of the correct 1.25 billion years left the inferred break mass essentially unchanged.
So the movement of the break from cluster to cluster is not easily dismissed as an artifact of choosing different ages or correcting the data.
There is another important check.
The apparent relationship between break mass and age can be distorted by distance. More distant clusters are harder to observe at low stellar masses, which means the sample preferentially retains distant clusters whose break masses are high enough to be measured.
The researchers modeled that effect with a generalized additive model.
After correcting for the distance-dependent bias, the age-break relationship became stronger rather than disappearing.
What could make the stellar recipe change?
The researchers connect the observed pattern to a physical property of the clouds from which stars form: the speed of sound in the molecular gas.
This may sound like an odd quantity to connect to the birth masses of stars, but it has a physical role in the fragmentation of collapsing molecular clouds.
Theoretical work predicts that changing conditions such as the gas temperature can change the characteristic scale at which a cloud fragments into individual stars. In the model examined here, an increase in the effective sound speed shifts the IMF break toward higher stellar masses.
The study’s authors therefore interpret the changing break masses as evidence that the environments in which the clusters formed were not identical.
Older clusters, on average, formed at earlier times when the Milky Way’s star-forming molecular clouds are expected to have been hotter and more metal-poor, corresponding to higher effective sound speeds. That provides a possible physical explanation for why older clusters tend to show higher break masses.
The important word here is possible.
The observations are consistent with this environmental model, and previous simulations support the predicted connection between changing conditions and the IMF. But the researchers caution that the Milky Way’s actual history is complicated.
For example, the Milky Way appears to have maintained a comparatively constant star-formation rate over the relevant period, unlike the more rapid evolution expected for the more typical galaxy considered in the paper’s comparison. Because of that, the paper says substantially improved modeling would be needed to test the break-mass evolution quantitatively for the Milky Way.
The Milky Way did not make every cluster under the same conditions
The age trend is only part of the story.
Clusters of roughly the same age also show a substantial range of break masses.
That means the Milky Way’s star-forming environments were not identical even at a given point in cosmic history.
The researchers interpret this scatter as consistent with the fact that galaxies contain different star-forming regions with different physical conditions. Different molecular-cloud environments can have different effective sound speeds, which in turn can produce different IMF break masses.
This helps explain why some previous measurements of nearby stellar populations found an IMF consistent with the standard Kroupa form and a break around 0.4–0.5 solar masses.
If several regions with somewhat different IMFs are averaged together, their combined population can look surprisingly similar to a single IMF. The effect can be particularly strong when observations cover a relatively small volume containing fewer environments.
The Gaia observations used in this study allow stars to be divided into more distinct populations over a much larger region, roughly 100 to 300 parsecs, helping reveal variation that can disappear when populations are averaged together.
In other words, what once looked like one universal stellar recipe may partly have been the result of mixing together many different recipes.
The result survived repeated tests
The researchers also examined whether their conclusion depended on the particular quality cuts used to select the 110 clusters.
It did not appear to.
When individual cuts were removed one at a time, the positive correlation between break mass and cluster age remained statistically strong. The full main sample had a bias-corrected Spearman correlation of 0.449 with a p-value of 8.7 × 10⁻⁷.
The researchers then substantially loosened the quality requirements, expanding the sample to hundreds of clusters. The larger sample produced a comparable correlation strength, while the greater number of clusters increased the statistical significance to p = 10⁻¹³.
That test was important because it showed that the age-break relationship was not simply created by the researchers’ decision to retain only the highest-quality clusters.
The paper also acknowledges limitations in the underlying measurements.
For example, recent work suggests that the stellar isochrones used to estimate masses may introduce a bias at low stellar masses. If that happens, the precise break mass measured for an individual cluster could change. The authors argue that such an error would be systematic, meaning the exact quantitative values and evolution could shift, while the qualitative result of break-mass variation should remain.
Unresolved binary stars are another complication. The analysis applies a correction for them, but the paper notes that their proper treatment remains unsettled and can significantly affect inferred mass functions, particularly at low mass.
These caveats do not erase the observed pattern, but they are part of why the researchers distinguish their strong evidence for variation from a complete description of every process controlling the IMF.
A universal IMF may be only a local approximation
The study was published in The Astrophysical Journal Letters, where the authors argue that the Gaia observations provide direct evidence against a fixed, universal IMF.
The significance reaches beyond the question of how Milky Way clusters formed.
Astronomers routinely use assumptions about the IMF when interpreting galaxies whose individual stars cannot be resolved. The assumed distribution of stellar masses affects estimates of quantities such as a galaxy’s total stellar mass and star-formation rate.
If the IMF changes with environment, those estimates can become biased when a fixed IMF is assumed.
The paper argues that the effect could be particularly important for galaxies observed at very high redshift by the James Webb Space Telescope, because the physical conditions in those early galaxies differ most strongly from those in the Milky Way today. Within the framework discussed by the authors, stellar masses and star-formation rates inferred under a fixed IMF could be significantly overestimated, potentially helping reconcile some apparently extreme observations with standard astrophysical and cosmological models.
But there is a deeper complication.
A galaxy does not form all of its stars under one set of conditions.
It builds its stellar population through countless star-formation events occurring across different places and times. If the IMF changes from one environment to another, the combined population of the entire galaxy becomes a mixture of those different IMFs.
The resulting aggregate distribution is not necessarily identical to the IMF of any one region, and it is not even guaranteed to have the same mathematical form as the family of IMFs used to describe its individual components.
That makes the idea of a single universal IMF increasingly difficult to maintain as a literal description of star formation.
The slopes remain an open question
The study does not show that every aspect of the IMF varies.
In fact, the researchers emphasize that they have demonstrated something more specific: strong evidence for variation in the break mass.
Whether the IMF’s power-law slopes themselves are universal remains much harder to determine.
That is because dynamical evolution affects those slopes much more strongly than it affects the break mass. A cluster can lose low-mass stars over time, changing its present-day slopes in ways that can resemble a change in the IMF itself.
Separating those effects would require clusters whose dynamical histories can be modeled accurately enough that any remaining slope changes could confidently be attributed to the IMF.
The authors also note that other physical effects may influence the IMF. Previous studies have reported evidence for a more top-heavy high-mass IMF under conditions such as low metallicity and high star-formation rates, while more recent work has found possible metallicity-dependent changes in the intermediate-mass slope.
The present study does not resolve those questions.
Instead, it establishes a particularly clean observational result: the characteristic break in the stellar mass distribution varies among Milky Way open clusters and changes systematically with cluster age.
And that brings the story back to the stars themselves.
For generations, the IMF has been treated as one of the basic ingredients that astronomers can assume when translating the light from a galaxy into a picture of its stars.
The new Gaia-based evidence suggests that the ingredient may not be fixed.
The stars born in the Milky Way appear to carry a record of the environments that made them—and that record changes with both place and time. The study’s observations are consistent with a universe in which the IMF is not one immutable recipe, but an environmental property of star formation.
The authors therefore argue that a universal IMF should no longer be treated as the default description of star formation, but as a local limit of a broader model in which the conditions of a star-forming environment help determine what kinds of stars are born.
