Resonant or asymmetric: The status of sub-GeV dark matter

Sowmiya Balan, Csaba Balázs, Torsten Bringmann, Christopher V. Cappiello, Riccardo Catena, Timón Emken, Tomás E. Gonzalo, Taylor R. Gray, Will Handley, Quan Huynh, Felix Kahlhoefer, Aaron C. Vincent

TL;DR

This paper provides the most comprehensive "health check" to date for sub-GeV dark matter models. By performing a global statistical analysis of all existing cosmological, astrophysical, and laboratory constraints, the authors demonstrate that simple symmetric fermionic dark matter is increasingly "cornered," while models featuring particle-antiparticle asymmetry or scalar particles remain highly viable and offer exciting targets for upcoming experiments.

What Problem Does This Paper Address?

For decades, the search for dark matter (DM) focused on WIMPs (Weakly Interacting Massive Particles) at the GeV-TeV scale. As those searches come up empty, the physics community has shifted focus toward "sub-GeV" dark matter. These lighter particles evade standard direct detection limits but face a different gauntlet of constraints: they can mess up the Big Bang's chemistry (BBN), leave imprints on the Cosmic Microwave Background (CMB), or produce X-rays in the modern universe.

The challenge is that most studies look at these constraints in isolation or use overly simplified "benchmark" models. This paper addresses the need for a global analysis—combining every scrap of available data into a single statistical framework to see which models actually survive and where the best places are to look next.

Methods

The authors utilized the GAMBIT (Global and Modular BSM Inference Tool) framework to perform both frequentist (profile likelihood) and Bayesian (posterior probability) scans of the parameter space.

They modeled DM as either a Dirac fermion or a complex scalar, interacting with the Standard Model via a "dark photon" mediator. The key parameters varied included DM mass, dark photon mass, the kinetic mixing strength ($\kappa$), and the dark sector coupling ($g_{DM}$).

Crucially, they implemented several "state-of-the-art" likelihoods, including:

  • Cosmology: Relic density, BBN element abundances, and CMB energy injection.
  • Astrophysics: X-ray limits from INTEGRAL/NuSTAR and self-interaction constraints from the Bullet Cluster.
  • Accelerators: Missing energy/momentum searches from NA64 and BaBar, and beam-dump scattering from LSND and MiniBooNE.
  • Direct Detection: Electron and nuclear recoil data from SENSEI, PandaX-4T, and others.

Datasets

The study synthesizes data from a vast array of experimental collaborations:

  • Cosmological/Astrophysical: Planck 2018, INTEGRAL, NuSTAR, XMM-Newton, and Bullet Cluster observations.
  • Laboratory: NA64, BaBar, LSND, MiniBooNE, SENSEI, DAMIC-M, DarkSide-50, XENON1T, and PandaX-4T.
  • New Benchmark: The authors introduce BP3 ($m_{A'} = 2.5 m_{DM}$, $\alpha_D = 0.3$), arguing that the traditional benchmarks (BP1 and BP2) used by the community are already in significant tension with existing data.

Key Insights and Ideas

  • The "Fine-Tuning" Penalty: Symmetric fermionic DM is in trouble. To avoid over-annihilating and ruining the CMB, it must live almost exactly on a "resonance" (where the dark photon mass is roughly twice the DM mass). Bayesian analysis reveals this requires extreme fine-tuning, making the model statistically less "natural."
  • Asymmetry as a Shield: If dark matter has an asymmetry (more particles than antiparticles, similar to normal matter), the annihilation rate in the late universe drops precipitously. This "unlocks" huge regions of parameter space that were previously thought to be ruled out.
  • Scalars are "Born Safe": Scalar DM annihilation is "p-wave suppressed," meaning it naturally slows down as the universe cools. This allows it to satisfy CMB constraints easily without needing an asymmetry or a perfectly tuned resonance.
  • The Bullet Cluster Matters: For asymmetric DM, the couplings can be large enough that DM particles start hitting each other. The authors found that Bullet Cluster data actually provides a meaningful "ceiling" on how large these couplings can be.

Key Results

  • Model Preference: Using Bayesian model comparison, the authors found a "strong" preference (Bayes factor of ~15) for asymmetric fermionic DM over the symmetric version.
  • Benchmark Tension: The most commonly used benchmark point (BP1) is excluded at over 95% confidence for both scalar and fermionic models.
  • Discovery Potential: The proposed LDMX experiment is identified as a powerhouse; it is projected to probe 64% of the currently allowed Bayesian posterior volume for these models.
  • New Target: The paper provides a new, robust benchmark point (BP3) that is consistent with all current data and serves as a clear target for the next decade of dark matter searches.

Limitations and Future Work

The authors acknowledge significant theoretical uncertainties in calculating the DM relic density near very narrow resonances. While they argue their global results are robust against these shifts, a more precise treatment of "kinetic decoupling" would be a valuable future improvement.

Additionally, the study focuses on the "invisible" decay of the dark photon ($m_{A'} > 2m_{DM}$). Future work could expand this to "secluded" scenarios where the dark photon is lighter than the dark matter, which involves a different set of experimental signatures.

Summary updated

Abstract

Sub-GeV dark matter (DM) particles produced via thermal freeze-out evade many of the strong constraints on heavier DM candidates but at the same time face a multitude of new constraints from laboratory experiments, astrophysical observations and cosmological data. In this work we combine all of these constraints in order to perform frequentist and Bayesian global analyses of fermionic and scalar sub-GeV DM coupled to a dark photon with kinetic mixing. For fermionic DM, we find viable parameter regions close to the dark photon resonance, which expand significantly when including a particle-antiparticle asymmetry. For scalar DM, the velocity-dependent annihilation cross section evades the strongest constraints even in the symmetric case. Using Bayesian model comparison, we show that both asymmetric fermionic DM and symmetric scalar DM are preferred over symmetric fermionic DM due to the reduced fine-tuning penalty. Finally, we explore the discovery prospects of near-future experiments both in the full parameter space and for specific benchmark points. We find that the most commonly used benchmark scenarios are already in tension with existing constraints and propose a new benchmark point that can be targeted with future searches.
BibTeX
@article{balan2025resonant,
  author = {Sowmiya Balan and Csaba Balázs and Torsten Bringmann and Christopher V. Cappiello and Riccardo Catena and Timón Emken and Tomás E. Gonzalo and Taylor R. Gray and Will Handley and Quan Huynh and Felix Kahlhoefer and Aaron C. Vincent},
  title = {{Resonant or asymmetric: The status of sub-GeV dark matter}},
  journal = {Journal of Cosmology and Astroparticle Physics},
  year = {2025},
  doi = {10.1088/1475-7516/2025/01/053},
}
Figure