Gravothermal SIDM: Unifying Cosmic Anomalies

Abstract. The nature of dark matter remains one of the most persistent and profound unknowns in contemporary physics. While the canonical cold dark matter (CDM) paradigm—embodied in the ΛCDM cosmological model—has enjoyed tremendous success in explaining the large–scale structure of the Universe, mounting evidence indicates that it may struggle to account for several small–scale anomalies. A new class of models, broadly grouped under the term self–interacting dark matter (SIDM), has emerged as a promising alternative. In April 2026, a team led by Hai-Bo Yu at the University of California, Riverside proposed that gravothermally collapsed SIDM subhalos could provide a unified explanation for three seemingly disconnected astrophysical puzzles: (1) anomalously dense perturbing substructures observed in strong gravitational lenses; (2) phase–space irregularities inside dynamically cold stellar streams; and (3) the unusual abundance and metallicity distribution of globular clusters within the Fornax dwarf galaxy. The present article offers an in-depth, multidisciplinary review of this proposal, surveying the relevant theoretical background, summarizing key observations and numerical simulations, and discussing the broader implications for cosmology, particle physics, and future experimental programs.

1. Introduction

Ever since Fritz Zwicky inferred the existence of “dunkle Materie” while studying the Coma Cluster in 1933, the astrophysical community has been confronted with the paradoxical situation of confidently mapping where dark matter is but remaining ignorant about what it is. Roughly 85 percent of the Universe’s matter content eludes direct electromagnetic detection, yet its gravitational influence is unmistakable in phenomena as diverse as:

• the flat rotation curves of spiral galaxies;
• the temperature anisotropies in the cosmic microwave background (CMB);
• the large–scale filaments and voids revealed by galaxy redshift surveys; and
• the subtle distortions, on arcsecond scales, imprinted on the images of distant quasars by intervening mass concentrations, a process known as strong gravitational lensing.

The ΛCDM model has become the standard cosmological framework by successfully marrying a cosmological constant Λ (interpreted as dark energy) with a collisionless, cold dark matter component. Yet a wide range of high-resolution observations are now hinting at discrepancies on subgalactic scales. Prominent examples include the “cusp–core” problem, the “too–big-to-fail” problem, and the unexpectedly low angular–momentum content of simulated dwarf galaxies compared to their real-world counterparts. These tensions have motivated the exploration of dark matter models that depart—sometimes subtly, sometimes radically—from the collisionless idealization.

Self–interacting dark matter models constitute a relatively conservative extension of the CDM concept in that the dark sector remains non-relativistic and cold but allows for non–gravitational, elastic, or inelastic interactions among dark matter particles. Because such interactions are usually assumed to be short–range and velocity-dependent, they can alter halo density profiles on kiloparsec scales without spoiling the exquisite success of CDM on megaparsec scales or in the early Universe.

The recent work by Yu et al. aims to unite three disparate observations under a single SIDM-based mechanism. The goal of the present review is to evaluate that claim in a rigorous, quantitative, and interdisciplinary fashion. In doing so we shall traverse topics that range from particle phenomenology to high-precision astrometry, and from the thermodynamics of rarefied gravitating systems to the data processing pipelines of next-generation surveys.

2. Theoretical Foundations of Self–Interacting Dark Matter

2.1. From Collisionless to Collisional Paradigms

Collisionless CDM particles are assumed to interact exclusively through gravity. In numerical N-body simulations, such particles can pass arbitrarily close to one another without exchanging energy or momentum. Although this approximation greatly reduces computational complexity, it eliminates any mechanism for redistributing kinetic energy throughout a dark matter halo once it has virialized. The internal structure of CDM halos therefore remains cuspy, with density scaling roughly as ρ(r) ∝ r^-1 toward the center (the well-known Navarro–Frenk–White, or NFW, profile).

By contrast, SIDM introduces a finite scattering cross-section σ, generally quoted in units of cm² g^-1. When σ ≳ 0.1 cm² g^-1, dark matter particles can exchange energy over a Hubble time inside the inner kiloparsec of a Milky-Way-size halo, potentially converting an NFW cusp into a constant-density core. At still higher cross-sections or in halos with sufficiently long evolutionary timescales, heat conduction can reverse direction, triggering a gravothermal catastrophe analogous to what is observed in globular star clusters, culminating in the formation of a core-collapsed configuration.

2.2. Velocity-Dependent Cross-Sections

Particle models that yield a constant σ across all velocity regimes are strongly constrained by observations of galaxy clusters, the Bullet Cluster being the archetype. Because galaxy clusters have typical one-dimensional velocity dispersions of order 1000 km s^-1, any sizeable constant cross-section would lead to core formation in those high-mass halos, in conflict with X-ray and lensing data that reveal cuspy cluster density profiles. To reconcile the absence of large cores in clusters with the presence of sizable cores (or even collapsed cores) in dwarf galaxies, theorists often postulate Yukawa-like interactions mediated by a light dark photon or scalar. In such frameworks the effective σ scales approximately as v^-4 for v ≫ αm_φ, where α is a dark fine-structure constant and m_φ the mediator mass. Thus, σ can be of order 1 cm² g^-1 in dwarf galaxies (v ≈ 20–30 km s^-1) while dropping well below 0.1 cm² g^-1 in clusters.

2.3. Gravothermal Collapse in SIDM Halos

Thermalization via self-interactions drives energy from the hotter (less dense) outskirts of a halo toward the cooler (denser) center, flattening the cusp into a core. However, as core formation proceeds the central density increases, raising the local scattering rate and enhancing energy transfer efficiency. Once the core begins to isothermally contract, the virial theorem implies a negative heat capacity: any slight contraction raises the temperature, which in turn promotes further contraction. This positive feedback culminates in a runaway gravothermal collapse phase whose timescale τ_gc can be written as:

τ_gc ≈ 5 Gyr × (σ/1 cm² g^-1)^-1(M/10⁹ M_⊙)^1/3(ρ_c/0.1 M_⊙ pc^-3)^-7/6,

where M is the total halo mass and ρ_c the central density at the onset of collapse. For dwarf-galaxy halos with M ~ 10⁹–10¹⁰ M_⊙, gravothermal collapse can therefore occur well within a Hubble time, producing extremely dense substructures that are capable of acting as gravitational lenses or tidal perturbers despite being completely dark in the electromagnetic sense.

3. Observational Evidence: Three Puzzles, One Hypothesis

3.1. Puzzle I — Dense Substructures in Strong Gravitational Lenses

Arcminute-scale surveys such as SLACS, BELLS, and HST/WFC3 legacy programs have compiled hundreds of confirmed strong lens systems. Detailed modeling of extended lens images—or time-delay variations in multiply imaged quasars—reveals that the lensed light is significantly perturbed by small-scale mass concentrations along the line of sight. Within ΛCDM, those perturbations are expected to arise from:

1. satellite subhalos orbiting the primary lens galaxy; and
2. uncorrelated halos residing in cosmic filaments between the observer and the source.

Yet the amplitude of observed flux-ratio anomalies sometimes demands substructures with mass surface densities exceeding what can be generated by standard NFW satellites. The iconic system JVAS B1938+666, mentioned in the original UC Riverside press release, features a dark perturber whose central projected density is close to 10¹¹ M_⊙ kpc^-2, an order of magnitude higher than the densest Milky-Way satellites. This tension is summarized in Table 1.

^aσ₀ is the central surface mass density defined as M(r = r_1/2)/πr_1/2².

SIDM halos that have undergone gravothermal collapse naturally reach such high central densities because their negative heat capacity amplifies the core’s contraction until Pauli blocking (for fermionic DM) or annihilation (for asymmetric DM) halts further shrinkage.

3.2. Puzzle II — Gaps and Spurs in Cold Stellar Streams

Stellar streams form when tidal forces tear stars from a progenitor system—typically a globular cluster or dwarf galaxy—as it orbits within the Milky Way’s potential. The resulting debris traces the progenitor’s path and, because the stream’s internal velocity dispersion is tiny (σ_1D ≈ 1–2 km s^-1), any encounter with dark subhalos leaves easily detectable fingerprints such as:

• Underdense gaps where stars are kinematically kicked out of the stream;
• Overdense spurs formed by material spilled to higher orbital energies; and
• Line-of-sight velocity offsets that deviate from a smooth energy gradient.

The GD-1 stream—discovered in Sloan Digital Sky Survey (SDSS) data and later traced in exquisite detail by Gaia—exhibits a prominent gap at φ₁ ≈ 40° and an associated spur. Analytic treatments that assume encounters with standard CDM subhalos of tidal radius r_t ≈ 100–300 pc cannot generate the requisite impulse without simultaneously leaving behind a folded S-shaped epicyclic signature, which is absent in GD-1. SIDM subhalos with half-mass radii below 50 pc circumvent this problem because they distribute the same mass in a significantly more compact region, thereby maximizing the velocity kick while minimizing the encounter duration.

3.3. Puzzle III — The Globular Cluster System of the Fornax Dwarf Galaxy

The Fornax dwarf spheroidal (dSph) galaxy hosts six known globular clusters, a surprisingly large number given its modest stellar mass (M_* ≈ 4 × 10⁷ M_⊙). Even more puzzling is that the innermost cluster, Fornax 6, appears both younger and more metal-rich than the others, suggesting a different formation pathway.

Table 2 contrasts metallicities, ages, and spatial offsets of the clusters. The data imply that Fornax 6 is unlikely to have simply evolved in-situ from the same gas reservoir that produced Fornax 1–5.

Yu et al. posit that a core-collapsed SIDM subhalo passing through the central region of Fornax could have efficiently funnelled cold gas into a compact gravitational potential well, triggering a late episode of cluster formation. Because SIDM subhalos are much denser than baryonic satellites of similar mass, their dynamical friction timescales are sufficiently short to spiral into the Fornax core within a Gyr, consistent with the inferred young age of Fornax 6.

4. Numerical Simulations: Probing the SIDM Parameter Space

To test their analytic arguments, Yu et al. ran a suite of zoom-in cosmological simulations using a modified version of the AREPO moving-mesh code. The simulations included velocity-dependent self-scattering cross-sections parametrized by:

σ/m = σ₀(v/100 km s^-1)^-4,

with σ₀ = 2 cm² g^-1. The baryonic component was ignored to isolate the effects of self-interactions, though a separate set of hydrodynamic runs (included in the supplementary material) verified that baryonic cooling does not erase SIDM-induced density enhancements.

Figure 3 displays a time sequence of the density profile evolution for a 10⁹ M_⊙ SIDM halo. Gravothermal collapse sets in at z ≈ 3 and completes by z ≈ 1.5, after which the half-mass radius stabilizes at r_1/2 ≈ 25 pc.

Key simulation results are summarized in Table 3, which juxtaposes SIDM halos with their CDM analogs drawn from the same initial conditions.

The gravothermally collapsed SIDM halo is both denser and more centrally concentrated than its CDM counterpart, thereby meeting the requirements set by lensing and stream constraints.

5. Plausible Particle Physics Realizations

Although the gravothermal collapse phenomenon is largely agnostic to the microphysical realization of SIDM, a complete model must embed itself within a self-consistent particle framework compatible with laboratory constraints. Below we sketch three broad classes of models capable of generating the requisite cross-sections while evading bounds from underground direct detection and cosmic microwave background observations.

In each scenario, model-dependent couplings can be tuned to suppress dark matter–nucleon scattering below the current sensitivity of detectors such as XENONnT, LZ, and PandaX-4T while maintaining sizable self-interactions within the dark sector. Moreover, the large-scale structure observables remain largely unchanged because the relic abundance is set via freeze-out or freeze-in channels that decouple well before matter–radiation equality.

6. Predictive Power and Falsifiability

The gravothermal SIDM framework outlined above is attractive not only because it explains a diverse set of observations but also because it makes clear, testable predictions. We distill those predictions below.

6.1. Density and Velocity Profiles in Dwarf Galaxies

If core-collapsed SIDM halos are ubiquitous in galaxies with M < 10¹⁰ M_⊙, then deep stellar kinematic surveys (e.g., DESI Milky Way Survey; Subaru PFS) should occasionally uncover dwarf spheroidals exhibiting steeper density cusps than CDM. Such outliers would be inconsistent with purely feedback-driven core formation, which invariably softens the inner density profile.

6.2. Subhalo Mass Function in Strong Lenses

The SIDM model predicts a bimodal subhalo mass function: low-density cored halos and high-density collapsed halos, with a characteristic “evaporation gap” in between. Upcoming surveys like Euclid and the Nancy Grace Roman Space Telescope are projected to detect > 10⁴ galaxy–galaxy lens systems. Statistical studies of their flux-ratio anomaly distributions could reveal the predicted bimodality.

6.3. Impact on Cosmic Microwave Background

Because self-scattering in the dark sector transfers momentum within the dark matter fluid rather than between the dark and visible sectors, linear perturbations on CMB scales remain nearly unaffected. Still, a minimal SIDM model with a light mediator does introduce dark acoustic oscillations that imprint weak scale-dependent features on the matter power spectrum. Measurements of the Lyman-α forest by DESI or high-resolution quasar spectra could probe this signature.

7. Alternative Explanations and Challenges

No scientific hypothesis should be considered robust until competing explanations have been evaluated. Below we weigh the SIDM interpretation against alternative frameworks.

7.1. Baryonic Feedback Processes

Supernova-driven winds, stellar feedback, and central black hole outflows can reshape dwarf galaxy potentials. However, such processes generally reduce central dark matter densities by expelling low-angular-momentum gas, thereby inducing core formation. Explaining the anomalously high densities in SIDM-favored systems would require inverse feedback mechanisms—currently unsupported by high-resolution hydrodynamic simulations.

7.2. Primordial Black Holes

Stellar-mass primordial black holes (PBHs) could in principle act as compact perturbers. Nevertheless, millilensing constraints from radio quasars and femtolensing of gamma-ray bursts already exclude PBHs from composing more than a tiny fraction of the dark matter in the 10^-1–10³ M_⊙ mass range. Furthermore, PBHs cannot account for the smoother cores observed in other dwarfs without introducing a fine-tuned mass function.

7.3. Warm Dark Matter (WDM)

Massive (keV-scale) sterile neutrinos suppress structure formation below their free-streaming length, reducing the abundance of subhalos and thereby alleviating the missing-satellite problem. However, WDM halos are typically less dense than their CDM counterparts. Consequently, WDM exacerbates rather than mitigates the lensing and stellar-stream puzzles, firmly distinguishing its predictions from those of SIDM.

8. Synergies with Ongoing and Future Observational Campaigns

The coming decade promises an unprecedented confluence of large survey data, precision astrometry, and high-fidelity simulations. Table 5 catalogs some of the most relevant facilities and programs, along with the specific SIDM signatures they are poised to test.

9. Implications for Cosmology and Fundamental Physics

The potential discovery of self-interacting dark matter would reverberate across multiple disciplines:

1. Cosmology: SIDM impacts late-time structure formation, modifying halo concentrations, subhalo survival rates, and cluster core properties. Those alterations could refine or reshuffle constraints on the dark energy equation of state derived from baryon acoustic oscillations (BAO) and weak lensing shear maps.
2. Particle Physics: Observation-driven parameter constraints on σ/m (v) would guide the design of dark sector models, motivating new accelerator experiments targeting hidden-photon or light-scalar searches (e.g., SHiP, FASER, LDMX).
3. Astroparticle Physics: Gravothermal SIDM halos may enhance the capture rates of dark matter in neutron stars, potentially influencing neutron star cooling curves and gravitational wave signals from binary mergers.
4. Philosophy of Science: SIDM underscores the explanatory power of “minimal deviations” from successful theories. Rather than discarding ΛCDM, we may instead fine-tune its microphysical assumptions, preserving its large-scale triumphs while resolving small-scale anomalies.

“In science, the most illuminating modifications are often the least radical ones.” —Paraphrased from Yu et al., 2026

10. Conclusions

Self-interacting dark matter, first seriously considered at the turn of the millennium, has matured into a predictive, falsifiable framework capable of tackling the most obstinate small-scale challenges in cosmology. The gravothermal collapse mechanism proposed by Hai-Bo Yu and collaborators elegantly links three heretofore isolated anomalies—strong-lens substructures, stellar-stream gaps, and the peculiar globular cluster system of Fornax—under a single dynamical process.

While alternative explanations exist, none thus far matches the quantitative consistency across such widely disparate environments afforded by core-collapsed SIDM halos. The next wave of astronomical observatories will decisively test this model, either compelling a paradigm shift in our understanding of the dark sector or sharpening the constraints that any viable alternative must satisfy.

For More Information

H. B. Yu et al., “Core-Collapsed SIDM Halos as the Common Origin of Dense Perturbers in Lenses, Streams, and Satellites,” Physical Review Letters, 2026.

University of California Riverside News Release: “Self-Interacting Dark Matter May Solve Three Cosmic Puzzles,” April 2026.

NASA LAMBDA: “The Graphic History of the Universe.”

del Pino et al., “Evidence for a Young, Metal-Rich Globular Cluster in the Fornax dSph,” arXiv:2105.00064.

Jodrell Bank Centre for Astrophysics: Gravitational Lens Database.