🔭 astrophysics

Modeling Galactic Rotation Curves to Map Dark Matter Distribution in the Galaxy: A Comparative Bayesian Analysis of Cuspy and Cored Halo Profiles Across the Sparc Database

This study utilizes Bayesian analysis with Gaussian Process Regression on the SPARC database to demonstrate that the observed preference for cored dark matter halo profiles over the cuspy NFW prediction is largely an artifact of unmodelled correlated data uncertainties, suggesting that resolving the Cusp-Core problem requires higher-resolution three-dimensional kinematic modeling rather than improved one-dimensional curve fitting.

Original authors: Hammed Adeniyi Lawal, Peter Taiye Samuel, Innocent E Davidson

Published 2026-10-20

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Hammed Adeniyi Lawal, Peter Taiye Samuel, Innocent E Davidson

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

For decades, astronomers have watched stars and gas clouds orbit the centers of spiral galaxies, expecting them to slow down as they move farther out, much like planets in our own solar system. Instead, they found that these outer regions race along at nearly the same speed as the inner ones. This unexpected flatness in the speed of rotation implies that galaxies are embedded in massive, invisible clouds of matter that we cannot see but can feel through their gravity. Scientists call this invisible substance dark matter. While its existence is widely accepted, a deep disagreement remains about what it looks like right at the center of a galaxy. Computer simulations based on our standard understanding of the universe predict that dark matter should pile up in a sharp, dense spike at the core. However, when astronomers measure the actual motion of stars and gas in real galaxies, the data often suggests a gentler, flatter distribution, like a smooth hill rather than a sharp peak. This mismatch between theory and observation has been a stubborn puzzle for over twenty years.

A new study by researchers in South Africa and Nigeria has taken a fresh look at this problem by re-examining the data from dozens of galaxies. The team analyzed a large collection of galaxy measurements known as the SPARC database, which contains detailed maps of how light and gas are distributed across many different types of galaxies. Their goal was to see if the data truly supports the idea of a smooth, flat center, or if the apparent evidence for this shape was actually an illusion created by how the data was analyzed. To do this, they first calculated the gravitational pull of all the visible matter in each galaxy—the stars in the disk, the central bulge, and the clouds of hydrogen gas. They then looked at the leftover motion that could only be explained by dark matter.

The researchers tested two competing ideas against this leftover motion. One idea used a model that predicts a sharp, dense spike of dark matter at the center, which is what standard computer simulations usually produce. The other idea used models that predict a smooth, flat core. When they used the standard way of fitting these models to the data, which assumes that every measurement point is independent of its neighbors, the smooth core models won almost every time. They fit the observed speeds of the stars and gas much better than the sharp spike models, especially in small, gas-rich dwarf galaxies. This result seemed to confirm that the universe's dark matter is distributed in a way that contradicts our best computer simulations.

However, the researchers suspected that the standard method of analysis might be missing something important. In reality, the measurements of galaxy rotation are not perfectly independent; errors in one part of the data often influence the measurements right next to it due to the way telescopes blur images and the complex, swirling motion of gas. To account for this, the team added a sophisticated statistical tool to their analysis that could detect and model these hidden connections between data points. When they re-ran their comparisons with this new tool included, the results changed dramatically. The clear advantage that the smooth core models had previously enjoyed largely disappeared. The sharp spike models, which had been rejected by the standard analysis, suddenly fit the data just as well as the smooth models.

This finding suggests that much of the historical evidence favoring a smooth, flat center for dark matter may not be a true reflection of the universe's structure, but rather a side effect of how the data was processed. The researchers conclude that the current measurements are not precise enough to definitively say whether dark matter forms a sharp spike or a smooth hill at the center of a galaxy. The apparent conflict between theory and observation may be an artifact of the noise in the data rather than a fundamental flaw in our understanding of dark matter. To solve this puzzle once and for all, the study argues that we need sharper instruments that can see these galaxies with greater clarity and new methods that can map the three-dimensional motion of gas, rather than relying on one-dimensional snapshots that hide the true complexity of these cosmic structures.

Technical Summary: Modeling Galactic Rotation Curves to Map Dark Matter Distribution in the Galaxy

Problem Statement
The study addresses the "Cusp-Core problem," a persistent discrepancy in astrophysics regarding the inner structure of dark matter halos. While Cold Dark Matter (ΛCDM) N-body simulations predict a steeply rising central density profile (a "cusp," specifically the Navarro-Frenk-White or NFW profile), observational data from spiral and irregular galaxies frequently favor a flatter, constant-density central region (a "core," such as the Pseudo-Isothermal profile). This discrepancy has remained unresolved for over two decades. A critical limitation in previous analyses is the standard assumption that uncertainties in rotation curve data points are statistically independent. In reality, factors such as beam smearing, non-circular gas motion, and inclination uncertainties introduce correlated systematic errors between adjacent radial bins. This paper investigates whether the historical preference for cored profiles over cuspy ones is a genuine physical signal or an artifact of unmodelled data correlations.

Methodology
The authors analyzed a sample of late-type galaxies drawn from the Spitzer Photometry and Accurate Rotation Curves (SPARC) database, supplemented by 21-cm HI radio data from the THINGS survey. The methodology involved a rigorous decomposition of galactic rotation curves into baryonic components (stellar disk, bulge, and gas) and a residual dark matter component.

Baryonic Decomposition: Stellar mass was estimated using 3.6 μm near-infrared photometry, applying Freeman-disk and Sérsic-bulge formalisms. Gas mass was derived from HI data, scaled to account for helium.
Halo Modeling: Two competing dark matter halo models were fitted to the residual velocity curves:
- Cuspy: The standard NFW profile derived from ΛCDM simulations.
- Cored: Pseudo-Isothermal (ISO), Burkert, and DC14 profiles.
Statistical Inference: The study employed Markov Chain Monte Carlo (MCMC) Bayesian inference to estimate parameters, including the stellar mass-to-light ratio ( $\Upsilon_*$ ), with appropriate priors.
Noise Modeling Comparison: The core innovation of the study was the comparison of two fitting regimes:
- Standard MCMC: Assumes independent, uncorrelated errors at each radial point.
- Gaussian Process Regression (GPR) Augmented MCMC: Incorporates a covariance kernel (Squared-Exponential/RBF) to model correlated systematic residuals directly from the data. This approach treats the noise structure as a set of hyperparameters ( $\sigma_{GP}$ and $l_{GP}$ ) sampled jointly with physical halo parameters.

Key Contributions

Integration of GPR into Halo Fitting: The paper introduces a Gaussian Process Regression extension to standard rotation curve fitting to explicitly account for correlated observational errors, moving beyond the assumption of independent noise.
Systematic Re-evaluation of the Cusp-Core Problem: By applying this new noise model to the SPARC database, the study isolates the extent to which the preference for cored profiles is driven by the physical halo structure versus unmodelled data correlations.
Case Study Analysis: Detailed kinematic decomposition of five benchmark galaxies (including the gas-dominated dwarf DDO 154 and massive spirals like NGC 2841) to demonstrate how the statistical preference for halo models shifts under different noise assumptions.

Results

Standard Assumption (Independent Errors): Under the traditional assumption of independent errors, cored profiles (specifically Pseudo-Isothermal and Burkert) significantly outperformed the NFW profile across most of the sample. For instance, in the dwarf galaxy DDO 154, the NFW profile yielded a reduced chi-square ( $\chi^2_\nu$ ) of 9.41, while the ISO profile achieved 2.40.
GPR-Augmented Analysis: When the GPR covariance term was included to model correlated errors, the statistical advantage of cored profiles diminished sharply. In DDO 154, the $\chi^2_\nu$ gap between ISO and NFW narrowed from 7.01 to 0.10 (0.94 vs. 1.04). Across the broader sample of 134 galaxies, the statistical preference for cored profiles over NFW, which existed in 88 galaxies under standard assumptions, was largely eliminated once correlations were modeled.
Mass-Dependent Trends: The study found that even in gas-dominated dwarfs like DDO 154, the clear preference for cored profiles observed under standard fitting largely disappears when GPR is applied, with both models converging to comparable goodness-of-fit values. In massive, bulge-dominated spirals, the data alone (even with GPR) remain insufficient to distinguish between cusp and core due to severe disk-halo degeneracy.
Convergence of Fit Quality: With GPR, both NFW and ISO profiles converged to comparable goodness-of-fit values ( $\chi^2_\nu \approx 1$ ) for most galaxies, suggesting that the "failure" of NFW in standard fits may be partly attributable to the mischaracterization of error structures.

Significance and Claims
The paper claims that a considerable portion of the historical observational evidence favoring cored dark matter halos may be attributable to unmodelled correlations in the data rather than a genuine absence of central cusps. The authors conclude that the Cusp-Core problem remains unresolved. They argue that the apparent discriminating power of standard one-dimensional rotation-curve fitting relies on an assumption (independent errors) that the data do not satisfy.

The study posits that resolving the Cusp-Core problem conclusively will likely require:

Higher-resolution instrumentation (e.g., the Square Kilometre Array) to reduce beam smearing and non-circular motion artifacts at the source.
Independent calibration of stellar mass-to-light ratios via vertical stellar velocity dispersion measurements to break disk-halo degeneracy.
A shift toward full three-dimensional kinematic modeling of interferometric data cubes, rather than relying on one-dimensional curve fitting.

The authors modestly assert that their results do not prove the existence of cusps, but rather demonstrate that current one-dimensional data and standard statistical methods are insufficient to rule them out. The findings align with independent recent analyses (e.g., Chase et al., 2026) that suggest unmodelled correlated errors are a significant systematic in halo inference.

More like this