Does the Milky Way's Dark Matter Halo Have a Core Instead of a Cusp?

A Galaxy That Spins Too Fast for Its Visible Stars

The Milky Way's outer stars move faster than its visible mass alone can explain. The mismatch, mapped in other spiral galaxies starting in the 1970s, helped establish the case for dark matter: an invisible mass that surrounds galaxies and dominates their gravity at large distances.

The Milky Way is difficult to study from the inside. Dust and gas obscure the disk, and measurements of how fast the Galaxy's stars and gas orbit at different distances have shown considerable scatter. But Gaia DR3 has enabled high-precision measurements of three-dimensional velocities and distances for large samples of stars, allowing researchers to reconstruct the Milky Way's rotation curve more precisely.

A rotation curve is a graph of orbital speed versus distance from the Galactic center. Together, the four Gaia-based compilations used in this study span about 5 to 30 kiloparsecs (kpc), the distance unit used in the paper. They show the rotation speed staying high and then declining beyond about 20 kpc. The question is what shape of dark matter halo can explain that curve.

The Difference Between a Cusp and a Core

The standard cold dark matter picture often uses a density profile called NFW, named after Navarro, Frenk, and White. In gravity-only computer simulations, dark matter halos tend to form a steep central cusp: density rises sharply toward the middle. Many observed galaxies, however, seem to prefer a core, where density flattens near the center.

This core-versus-cusp issue is a long-standing tension in small-scale cosmology because it tests how dark matter behaves on small scales. In the simplest gravity-only picture, simulations predict cusps. Alternative scenarios, such as self-interacting dark matter, can produce cores.

A new preprint by Aryan Singh and Shantanu Desai of IIT Hyderabad uses four recent Milky Way rotation-curve datasets built from Gaia DR3 and other surveys. They compare three models for the Galaxy's ordinary, or baryonic, matter; seven dark matter density profiles; and three MOND interpolating functions.

What Gaia's Stellar Motions Actually Test

The team did not directly observe dark matter. Instead, they asked which model best reproduces the observed orbital speeds. They used Bayesian model comparison, a statistical method that computes how probable the data are under each model after integrating over its parameters. They calculated Bayes factors relative to a baseline model combining each baryonic model with the NFW profile. In their scale, a Bayes factor above 100 counts as decisive evidence.

The four rotation-curve datasets are not identical. One uses 18 data points from red giant and OB stars. Another uses 34 points from red giant stars from APOGEE and LAMOST, with distances derived by combining Gaia DR3 and 2MASS information. A third uses 37 points from red giant stars using APOGEE DR17 spectroscopy and Gaia, WISE, and 2MASS photometry. A fourth uses 19 points reconstructed from three-dimensional stellar velocities. The datasets use different tracer stars, scale-height cuts, and treatments or corrections for asymmetric drift, a subtle drift caused by the random motions of stars. The paper labels them D1 through D4.

Einasto Wins Most, but Not Every, Round

The headline result is that the Einasto profile is the preferred phenomenological fit over NFW across most combinations of datasets and baryonic models. The Einasto profile is flexible: it can describe either a cusp or a core. Here, the best-fit shape parameter was positive in every baryonic-dataset combination, pointing toward a cored inner density rather than a steep cusp.

For D3 and D4, the evidence was especially strong: Bayes factors greater than 100 for all three baryonic models. For D2, the beta and Burkert profiles had even higher Bayes factors than Einasto, though Einasto still beat NFW decisively. D1 gave only substantial, not decisive, evidence for Einasto over NFW.

Cored profiles as a class performed well. The Burkert and beta profiles were strongly preferred over NFW in several datasets. The isothermal profile, by contrast, fit poorly. The generalized NFW profile gave mixed results, sometimes favored and sometimes disfavored. The Hernquist profile was favored over NFW across all datasets, though generally less strongly than Einasto.

MOND Fits Less Well in This Test

MOND, or Modified Newtonian Dynamics, proposes that gravity itself changes in the low-acceleration regime, potentially removing the need for dark matter. The authors tested three MOND interpolating functions, which describe the transition between ordinary and modified gravity. They ran them both with a fixed acceleration scale and with that scale allowed to vary.

Within these implementations, MOND provided poorer fits than the preferred dark matter profiles. The standard MOND paradigm was strongly disfavored compared with NFW across almost all datasets. One reason is that the MOND fits tend to produce flat or constant orbital speeds at large radii, while the Gaia-based rotation curves decline beyond about 20 kpc.

That does not settle MOND's fate. The analysis omits the external field effect, which arises when a galaxy feels the gravity of its larger environment, and it does not include non-spherical baryonic structure. Those omissions could affect the comparison. The paper is also a preprint, and its conclusions are model-dependent.

No Baryonic Model Takes the Crown

The researchers also compared the three baryonic models against one another while using the Einasto dark matter profile. None was decisively favored. Bayes factors were mostly below 10, meaning the data do not strongly prefer one treatment of the Milky Way's bulge and disk over another. The highest preference, for one baryonic model with one dataset, was only around 20.

That matters because the rotation curve is a sum of visible and dark contributions, and the paper notes that the disk mass was poorly constrained in some fits, especially when the dark halo dominates the outer rotation curve. That could make the Bayesian evidence sensitive to prior choices. However, the authors argue that because the same baryonic model is used when comparing dark matter profiles, the ranking among dark matter models should be fairly robust. They also checked narrower priors and found the Einasto preference held.

Where the Verdict Could Shift

The four datasets all depend on Gaia DR3, and three cover nearly the same radial range. They share systematic uncertainties, including the zero-point of parallax measurements and assumptions about the Sun's distance from the Galactic center and the Galaxy's circular speed. They are not truly independent tests.

Reconstructing a rotation curve also depends on the scale height of the tracer stars and on corrections for asymmetric drift. A separate study estimated systematic errors of about 6 percent inside 14 kpc, rising to perhaps 15 percent farther out in a worst-case scenario.

Bayesian model comparison does not prove that the Milky Way has a cored dark matter halo. It ranks models within a chosen framework and dataset. Still, the result adds to a growing body of evidence that the standard cuspy NFW picture may not be the best description of our Galaxy.

Upcoming facilities such as the Vera Rubin LSST, the Square Kilometer Array, the Thirty Meter Telescope, and the Extremely Large Telescope could provide more insights into the Galaxy's dark matter. For now, within these Gaia-based datasets and models, the Milky Way's invisible mass looks less sharply concentrated than many simulations predict, suggesting that the dark heart of our Galaxy may be gentler than expected.