Dark Halos of Galaxies: What Stellar Streams Reveal About the Shape of the Invisible
Stellar streams are used as probes of dark matter halo structure. A hierarchical Bayesian model, previously validated on simulations, is applied to 32 streams from the STRRINGS catalog. For each stream, assuming an axisymmetric halo, posterior distributions of the flattening parameter q are derived from the observed track projection. An extra dispersion term is introduced to account for track systematics and model mismatch, enabling the selection of a "golden" subsample of 17 highly informative streams. The population distribution, found via importance sampling, indicates a flattened halo shape for the golden subsample (μ_q≈0.72, σ_q≈0.34) and a nearly spherical one for streams dominated by extra dispersion. The result is consistent with hydrodynamic cosmological simulations. This work provides the first constraints on halo shape beyond the Local Group and sets the stage for analysis of future surveys from Euclid and LSST.
Context
Dark matter makes up about 85% of all matter in the Universe, yet its nature remains a mystery. Within the standard cosmological model of the Big Bang and dark energy, it forms vast halos around galaxies. However, the shape of these halos—whether oblate, prolate, or spherical—depends on many factors, including merger history and the influence of ordinary matter. Back in the 1930s, Fritz Zwicky suspected the presence of hidden mass in galaxy clusters, and in the 1970s, Vera Rubin proved its existence on the outskirts of spiral galaxies. Ever since Edwin Hubble discovered the expansion of the Universe and the world of galaxies, studying their structure has become a key to cosmology. Understanding the shape of the halo is a test of dark matter theories and galaxy formation models, as well as a bridge to unraveling its nature. Just as entropy tends to a maximum in an isolated system, dynamical relaxation rounds out initially triaxial halos, and the observed shape carries the imprint of this evolution.
Methods
Unlike the Milky Way, where streams can be studied with full 3D kinematics, for extragalactic systems only sky projections are available. The researchers applied direct modeling: for each of the 32 streams from the STRRINGS catalog, they generated artificial stellar tails in trial gravitational potentials, assuming an axisymmetric Navarro–Frenk–White model for the dark matter halo. By comparing the observed stream track with the model, they estimated the flattening parameter q using Bayesian inference. To account for systematic errors and model imperfection, an additional variance was introduced into the likelihood function, tuned by the data itself. This approach resembles gravitational lensing analysis, where the image shape reveals the mass distribution, but here the entire halo potential acts as the 'lens'. To increase the model's robustness, the particle ordering scheme along the stream was also modified.
Results
After individual fitting of 32 streams, scientists divided them into a 'gold' (17 objects) and 'bronze' (15 objects) sample based on how strongly the result depends on additional systematics. For the gold set, where the data are most informative, the population analysis yielded a mean dark halo flattening μ_q ≈ 0.72 with a spread σ_q ≈ 0.34. This means that the typical dark matter halo turned out to be slightly oblate (like a pumpkin) rather than prolate (like a melon) or perfectly round. The spherical case (q=1) is excluded at about 1.5σ. Interestingly, the 'bronze' sample showed an almost spherical distribution, but with much greater uncertainty, indicating a lack of information rather than a real shape. This result is the first measurement of dark matter halo shape for an entire population of galaxies outside the Local Group. Additionally, a trend of alignment between oblate halos and the galaxy disk was discovered, as expected from theoretical considerations.
Implications
The obtained mean flattening of about 0.72 is in good agreement with predictions from hydrodynamic cosmological simulations, such as IllustrisTNG and Auriga, which give a flattening of about 0.7 for the inner parts of halos. This confirms that baryonic physics (gas, stars, feedback) significantly rounds and flattens the initially triaxial dark matter halos formed in collisionless simulations. Discrepancies with earlier works that suggested sphericity may be explained by different analysis methods and samples. Importantly, the presence of oblateness affects the interpretation of direct dark matter detection experiments, as the local density depends on the halo shape.
Future development
Future sky surveys, such as Euclid and Rubin/LSST, will increase the number of known extragalactic streams by orders of magnitude. This will turn the statistical approach used in this work into a standard tool of cosmology. It is already clear that key improvements will include accounting for the baryonic disk in models, freeing the position of the stream progenitor, and transitioning to triaxial potentials. Moreover, incorporating information on the width and surface brightness of streams, not just their trajectories, will dramatically strengthen constraints. The development of machine learning methods for automatic stream extraction in deep images will also accelerate the accumulation of statistics. In the future, this will allow mapping the distribution of dark matter in the local Universe with unprecedented detail, testing alternative theories of gravity.
Impact
The results will impact several areas: extragalactic astronomy (understanding galaxy formation), particle physics (constraints on dark matter models), and cosmology (tests of the standard cosmological model).
Next steps
Next steps include a reanalysis with explicit inclusion of the stellar disk in the potential model and applying the method to new streams discovered in deep surveys to refine dark halo parameters.
Key open problems
Connection to unsolved problems in physics: the shape of dark matter halos is directly linked to the nature of dark matter particles, their interaction with baryons, and the accuracy of the standard cosmological model. The discrepancy between predictions of purely collisionless simulations and observations indicates the need to incorporate complex feedback physics, which remains a challenge for modern galaxy simulations.
🎯 The longest stream in the STRRINGS catalog stretches almost 400 degrees across the sky—more than a full circle when viewed edge-on! Such giant arcs are ideal natural 'scales' for weighing dark matter.
Key numbers
- Mean halo flattening in gold sample: 0.72
- Intrinsic scatter of flattening: 0.34
- Number of streams in gold sample: 17
- Mean distance to galaxies: 60 Mpc
- Mean systematic variance: 3.12 kpc