Self-interacting dark matter deepens a galactic bar puzzle in simulations

A puzzle about galactic rotation

A stellar bar has a pattern speed: the angular rate at which its overall shape rotates. Through gravitational interactions, a bar can transfer angular momentum to the surrounding dark matter halo. This process, called dynamical friction, helps slow the bar.

Astronomers also characterize bars using a ratio called R. It compares the corotation radius, where the circular orbital angular speed matches the bar’s pattern speed, with the bar radius—half its full length. Bars with R below 1.4 are conventionally called “fast”; those above 1.4 are “slow.” These labels therefore describe a relationship between rotation and size, not simply an absolute rotation rate.

Cosmological simulations often produce larger R values than observations of nearby galaxies. This “fast-bar tension” does not mean all observed bars are fast: one observational sample discussed in the paper contains mostly slow bars, yet its average R remains substantially below a previously reported simulation value.

Three versions of the same simulated universe

Asiyeh Habibi and colleagues investigated whether self-interacting dark matter could reduce this discrepancy. Their study, posted to arXiv on November 6, 2026, is a preprint; the supplied material does not establish peer-reviewed publication.

They used AIDA-TNG simulations of a cosmic box 51.7 megaparsecs across. Three runs shared the same initial conditions and galaxy-formation prescriptions but used different dark matter physics. The standard cold dark matter model, CDM, was compared with SIDM1, which permits elastic scattering with a constant cross-section per unit mass of 1 square centimetre per gram, and vSIDM, whose scattering cross-section depends on particle speed.

The researchers selected disk galaxies containing more than 10 billion solar masses in stars at four redshifts: 2.0, 1.5, 1.0 and 0.2. Redshift identifies different stages of cosmic history, with the larger values corresponding to earlier epochs. The comparison stops at 0.2 because the SIDM1 run had not yet reached the present-day endpoint.

They identified bars by measuring a twofold pattern in the stellar mass distribution and estimated bar lengths using its Fourier components. Pattern speeds came from a method requiring the particles’ full three-dimensional positions and velocities. Unlike observational methods for distant galaxies, this approach can exploit information available directly inside a simulation.

The team compared both the full populations and matched galaxies: counterparts originating from the same initial conditions that were barred and had reliable measurements in all three runs.

Similar bar numbers, different dynamics

The fraction of disks hosting bars, measured as a function of stellar mass, was broadly similar across the models. Their rotational properties differed more clearly.

At redshift 0.2, the reported mean R values were 3.21 for CDM, 3.63 for vSIDM and 4.07 for SIDM1. All three means lie in the slow-bar regime. Statistical comparisons found both self-interacting models’ R distributions different from CDM at this epoch.

The strongest pattern-speed result concerned SIDM1. In the matched sample pooled across redshifts, its pattern-speed distribution differed significantly from CDM’s, whereas vSIDM’s did not. Thus, the evidence does not support treating both self-interacting models as equally distinct from standard dark matter in every comparison.

More than stronger braking

Between redshifts 2 and 0.2, average pattern speeds declined by about 61% in CDM, 68% in vSIDM and 69% in SIDM1. The authors describe this as a roughly 10% larger fractional decline in the self-interacting cases—not an additional 10 percentage points.

A possible contributor is the directionality of dark matter motion. In matched galaxies, halo densities and velocity dispersions were generally statistically similar, but self-interacting halos had less radially biased motions. The authors suggest these differences could alter angular-momentum exchange, directly and through the stronger bars found particularly in SIDM1. This is a proposed explanation, not an isolated causal demonstration.

Crucially, lower pattern speeds were already apparent at early sampled epochs. Stronger subsequent friction therefore cannot explain the whole difference. A connection to bar formation is plausible, but the study did not determine when individual bars formed.

What the result leaves open

The samples change between redshifts, so declining population averages do not prove that the same bars slowed by those amounts. Only five matched systems were available at redshift 2. The study also uses one cosmic volume and one galaxy-formation model; numerical resolution and exclusions of unreliable measurements remain limitations.

Bar size complicates interpretation too: the simulations fail to reproduce the observed increase of bar length with stellar mass, and R depends on that length as well as rotation.

For these two interaction prescriptions, self-interacting dark matter does not solve the fast-bar tension. That is a constraint on these simulations, not a verdict on every possible self-interacting dark matter model.