Boosted Dark Matter from Sagittarius A$^\star$

Abstract: It was recently demonstrated that black hole binaries can gravitationally accelerate ambient dark matter (DM), producing a continuous flux of particles with velocities far exceeding those of the galactic halo. We extend this analysis to the Milky Way's nuclear star cluster, where stellar-mass black holes are expected to orbit in close proximity to the supermassive black hole Sagittarius A$<sup>\star$. Using numerical simulations, we compute the flux of gravitationally-boosted DM sourced by this region. Because of the high DM density and large population of black holes orbiting deep within Sagittarius A$<sup>\star$'s gravitational potential, the resulting DM ejecta attain substantially higher rates and energies compared to galactic black hole binaries, with simulated particles reaching velocities of up to $\sim 25,!000 \, \rm km/s$. We find that the nuclear star cluster is therefore the dominant source of gravitationally-boosted DM in the Milky Way. Even under conservative assumptions about the DM profile in the inner galaxy, the ejected DM flux from this region can render large-volume DM detectors competitive with lower-threshold experiments in the sub-GeV mass range, independently of the underlying DM particle model. The gravitational nature of the boost also opens up a sizable detection window into heavy inelastic DM scenarios that are otherwise largely inaccessible to conventional halo DM searches.

• The paper establishes that stellar-mass black holes orbiting Sagittarius A* are the Milky Way’s dominant source of gravitationally boosted dark matter, producing velocities up to approximately 25,000 km/s through three-body encounters.
• The simulations find that only a few dozen black holes on tight, high-eccentricity orbits dominate the highest-energy tail, while the predicted flux remains sensitive to black-hole demographics, the inner dark-matter profile, and possible intermediate-mass black holes.
• The results show that LZ, XENONnT, PandaX-4T, and DarkSide-50 can probe elastic dark matter down to roughly 30–60 MeV and TeV-scale inelastic dark matter with mass splittings up to about 3–9 MeV.

Overview

This paper by Acevedo and Ritz (2606.30724) extends the recently developed framework of gravitationally boosted dark matter (DM) to the Milky Way's nuclear star cluster (NSC). Building on the earlier demonstration that black hole binaries can eject transiting halo DM via the gravitational slingshot mechanism (Acevedo et al., 9 Mar 2026), the authors show that stellar-mass black holes orbiting the supermassive black hole Sagittarius A$^\star$ (Sgr. A$^\star$) constitute the dominant galactic source of boosted DM. The central claim is strong: even under conservative assumptions about the inner-galaxy DM profile, the resulting flux renders large-volume detectors (LZ, PandaX-4T, XENONnT, DarkSide-50) competitive with lower-threshold experiments in the sub-GeV regime, and opens a detection window into heavy inelastic DM with MeV-scale splittings that is otherwise inaccessible.

The key physical advantage over galactic binaries is threefold: roughly $2.5 \times 10^5$ stellar-mass black holes are expected within $\sim 1.5$ pc of Sgr. A$^\star$; these objects move at much larger orbital speeds while maintaining large separations from the primary; and the region hosts the highest DM density in the Galaxy. The simulated ejecta reach velocities up to $\sim 25{,}000\ \rm km/s$, far above the Standard Halo Model (SHM) cutoff of $\sim 546\ \rm km/s$ — approximately 100 standard deviations above the SHM mean velocity.

Ejection dynamics

The analysis rests on the restricted three-body regime $m_\chi \ll M_2 \ll M_1$, where $$M_1 = M_{\rm Sgr A^*} \simeq 4 \times 10^6\,M_\odot$$ and $M_2$ is a stellar-mass black hole. In the secondary's rest frame, a close encounter is elastic two-body scattering with deflection angle set by $^\star$0. Transforming back to the barycenter frame yields an energy shift

$^\star$1

maximal for a head-on encounter with full back-deflection ($^\star$2, $^\star$3), giving $^\star$4. Two consequences follow directly. First, the kinematically allowed ejection energies depend only on orbital configuration, not on the secondary's mass; heavier scatterers simply eject more frequently, with the spectrum scaling as $^\star$5. This scaling is exploited to accelerate Monte Carlo convergence and underlies the paper's degeneracy argument between black hole number and mass. Second, because the energy gain scales with orbital speed, black holes on short-period, high-eccentricity orbits dominate the hard tail. For the most energetic simulated orbit ($^\star$6 AU, $^\star$7), the pericenter velocity reaches $^\star$8, implying an idealized maximum ejection speed of $^\star$9; numerically resolved ejecta reach less than half this value due to the vanishing probability of exact pericenter ejections.

A caveat noted explicitly: the point-mass treatment applies only to compact objects. Extended systems such as main-sequence stars impose a minimum impact parameter at their physical radius, precluding the strongest deflections, so the mechanism is specific to black holes (and other compact remnants).

Simulation methodology

The authors employ the Velocity Verlet Monte Carlo of Ref. (Acevedo et al., 9 Mar 2026), evolving ensembles of three-body systems and constructing the differential ejection rate as a product of an incoming flux factor and an ejection probability distribution. Several refinements are specific to the extreme mass ratio:

• Static primary: Sgr. A$2.5 \times 10^5$0 is held fixed at the origin; comparison against fully dynamical runs confirms no impact on the spectrum.
• Gravitational focusing corrections: since the simulation boundary lies well inside Sgr. A$2.5 \times 10^5$1's sphere of influence, the boundary density and dispersion are enhanced by factors $2.5 \times 10^5$2 relative to asymptotic values, derived via Liouville's theorem in an appendix.
• Isotropy approximation: ejection direction is taken isotropic and independent of energy, validated in prior work.
• Newtonian gravity only: justified by minimum pericenter distances of $2.5 \times 10^5$3 AU, far exceeding the gravitational radius; particles approaching either mass within $2.5 \times 10^5$4 its gravitational radius are removed (a negligible fraction).

The simulation produces a two-component spectrum: a low-energy component arising from soft scatterings whose normalization depends artificially on the simulation volume, and a convergent high-energy tail beginning near the halo velocity cutoff. Only the latter is used. This is an important methodological point — the physical signal resides entirely in the hard-scattering component.

Modeling the nuclear star cluster

Black hole population. Assuming spherical symmetry, steady state, and isotropy (justified by two-body relaxation times of $2.5 \times 10^5$5 yr, shorter than the NSC age), the phase-space distribution is parameterized as a power law in specific energy, $2.5 \times 10^5$6. The benchmark slope $2.5 \times 10^5$7 corresponds to the Bahcall–Wolf cusp, which the authors argue is conservative: shallower slopes underpopulate short-period orbits and suppress the highest-energy flux. The normalization adopts recent star-formation-history analyses yielding $2.5 \times 10^5$8 black holes of $2.5 \times 10^5$9 within $\sim 1.5$0 pc. Orbital element space is discretized into triangular bins in $\sim 1.5$1 variables, with successive zoom-in refinement of the shortest-period bin down to $\sim 1.5$2 AU; further subdivision changes the experimental reach only at $\sim 1.5$3 level.

A consistency check against the GRAVITY constraint on extended dark mass from S2's prograde precession ($\sim 1.5$4 at $\sim 1.5$5) shows the fiducial cusp implies $\sim 1.5$6 enclosed within S2's apocenter — marginally consistent, though the bound strictly applies to diffuse distributions and relaxes for granular mass. Notably, because the signal scales as $\sim 1.5$7, combinations of fewer, heavier black holes can reproduce or exceed the fiducial flux while remaining well below the enclosed-mass limit.

Dark matter profile. An NFW halo with local density $\sim 1.5$8 and inner slope varied over $\sim 1.5$9 to $^\star$0 brackets the uncertainty; the velocity dispersion is fixed at $^\star$1. Matching to Sgr. A$^\star$2's potential occurs at $^\star$3 pc, where the escape velocity equals the dispersion. Crucially, only unbound particles transiting the NSC are counted — under a standard NFW this accounts for merely $^\star$4 within S2's apocenter versus $^\star$5 from the full extrapolated profile. Including bound DM would require self-consistent depletion modeling beyond the ensemble approach, so the quoted fluxes are conservative lower estimates; no DM spike is invoked.

Flux and velocity distribution

Stacking all bins, the low-energy flux is dominated by the numerous black holes on long-period orbits, while the high-energy tail is controlled by a very small population: only about 30–40 black holes drive the flux at the highest resolvable energies. As a benchmark of population degeneracy, a single intermediate-mass black hole of $^\star$6 placed in the shortest-period bin produces a spectrum mildly suppressed at low energies but modestly enhanced at high energies — comparable to the entire stellar-mass population in the tail. Such an object remains allowed by astrometric constraints, which permit dark perturbers up to $^\star$7 at these orbital distances; whether one exists is an open observational question, though LISA may partially probe it.

The resulting velocity distribution at Earth extends to $^\star$8 with a much lower normalization than the SHM. Secondary effects — Sgr. A$^\star$9's negligible proper motion ($\sim 25{,}000\ \rm km/s$0), the $\sim 25{,}000\ \rm km/s$1 potential gradient along the baseline, and source evolution timescales — are each shown to be inconsequential for the spectrum.

Direct detection prospects

Rates are computed with the standard recoil integral using each experiment's reported ionization-only efficiency; when run with the SHM input, the procedure reproduces published limits to within a factor of 2–3, deemed adequate for relative comparisons. Sensitivity lines terminate at cross-sections $\sim 25{,}000\ \rm km/s$2, above which rock overburden attenuation invalidates the ballistic propagation assumption.

Interaction	Experiment	Extended reach
Elastic SI	Xenon detectors (LZ, XENONnT, PandaX-4T)	masses down to $\sim 25{,}000\ \rm km/s$3 MeV
Elastic SI	DarkSide-50 (argon)	masses down to $\sim 25{,}000\ \rm km/s$4 MeV
Elastic SD (neutron)	PandaX-4T	masses down to $\sim 25{,}000\ \rm km/s$5 MeV
Inelastic SI, $\sim 25{,}000\ \rm km/s$6 TeV	LZ	splittings up to $\sim 25{,}000\ \rm km/s$7 MeV

For spin-independent elastic scattering, DarkSide-50 significantly outperforms dedicated low-threshold experiments near 30 MeV thanks to its lighter target, while xenon detectors become competitive with CRESST and SuperCDMS depending on the assumed halo slope. For neutron-only spin-dependent scattering at PandaX-4T, the projected sensitivity surpasses current limits at the low-mass end if an intermediate-mass black hole is present. These gains are nearly model-independent: they derive from the gravitational boost alone, requiring no assumption about the DM–SM coupling beyond the scattering operator itself.

The inelastic analysis is the most distinctive result. Because the boost is gravitational, the energy gain carries no penalty at large $\sim 25{,}000\ \rm km/s$8, uniquely enabling TeV-scale endothermic scattering $\sim 25{,}000\ \rm km/s$9. For LZ, the boosted flux probes splittings up to $\sim 546\ \rm km/s$0 MeV in a largely model-independent manner (up to cross-sections $\sim 546\ \rm km/s$1, beyond which the $\sim 546\ \rm km/s$2 coherence scaling becomes model-dependent), extending to $\sim 546\ \rm km/s$3 MeV at larger cross-sections. The splitting reach is partly limited by the experiment's $\sim 546\ \rm km/s$4 keV maximum recoil acceptance rather than by kinematics, implying that a wider recoil window would directly improve it. The authors further observe that the maximum nuclear recoil for TeV DM at $\sim 546\ \rm km/s$5 on carbon exceeds $\sim 546\ \rm km/s$6 MeV, making large-volume neutrino detectors a plausible complementary probe for splittings $\sim 546\ \rm km/s$7 MeV; a detailed treatment is deferred to forthcoming work.

Limitations and open questions

Several assumptions bound the results. The black hole number is the largest uncertainty, depending sensitively on the assumed metallicity of the NSC; the fiducial $\sim 546\ \rm km/s$8 also sits marginally at the S2 extended-mass constraint, and the resolution relies on the granular-mass loophole. Mass segregation stronger than the Bahcall–Wolf cusp — which would enhance high-velocity ejections — is not modeled. The DM inventory excludes bound particles around Sgr. A$\sim 546\ \rm km/s$9, and no spikes are considered, so the flux should be viewed as a floor rather than an estimate. On the particle side, the analysis covers nucleus scattering only; the analogous electron-scattering extension requires detector response modeling and may overlap with stellar-cooling and cosmological constraints. Finally, the existence of an intermediate-mass black hole in the Galactic center remains unconfirmed, leaving open whether the enhanced high-energy sensitivity it provides is realized in nature.

Conclusion

This work establishes the nuclear star cluster as the dominant source of gravitationally boosted DM in the Milky Way, with resolved ejection velocities up to $m_\chi \ll M_2 \ll M_1$0 sourced largely by a few dozen black holes on the tightest orbits. The resulting flux extends large-volume detectors into sub-GeV elastic parameter space competitively with purpose-built low-threshold experiments, and — uniquely among boosting mechanisms, since it involves no non-gravitational acceleration — grants access to TeV-scale inelastic DM with MeV splittings previously untested by direct detection. The projections are deliberately conservative, counting only unbound transiting halo DM under a mild cusp, which suggests the quoted sensitivities understate the full potential of the mechanism.