Dark matter (DM) annihilation or decay in the neutral hydrogen of quiet dwarf galaxies produces Hα recombination emission; the first limits, derived by uniform averaging of MUSE observations of Leo T (Leane, Phys. Rev. Lett. 137, 071002 (2026)), already lead over parts of the eV–GeV DM mass range, and the source explicitly defers the optimized analysis. The open problem is to quantify and realize that optimization: (i) replace uniform aperture averaging with a matched filter weighted by the expected dark-matter morphology (ρDM2\rho_{\rm DM}^2 for annihilation, ρDM\rho_{\rm DM} for decay) and re-derive the flux and annihilation/decay limits with a quantified signal-to-noise gain; (ii) extend the extraction to Hβ and formalize the metal-line-forest test that distinguishes a dark-matter recombination signal from residual star formation or shocks; and (iii) forecast the reach of stacked samples of Leo-T-like dwarfs and of a realistic Extremely Large Telescope (ELT)/HARMONI observation that accounts for the instrument's smaller field of view against the extended dark-matter signal. No work implementing any of these components is known in the literature to date.

Background

Indirect detection of dark matter (DM) searches for the annihilation or decay products of DM particles rather than the particles themselves. For DM masses below a few GeV — the MeV-scale regime — gamma-ray satellites and nuclear-recoil direct detection lose leverage, so new observational channels matter. Leane (Phys. Rev. Lett. 137, 071002 (2026)) proposed such a channel: if DM annihilates (with rate ⟨σv⟩\langle\sigma v\rangle, weighted by the line-of-sight integral of the squared DM density, the JJ-factor) or decays (with lifetime τχ\tau_\chi, weighted by the density integral, the DD-factor) inside a dwarf galaxy dominated by neutral atomic hydrogen (H I), the injected photons or electrons ionize the gas, and the subsequent recombination cascade emits Hα photons — the hydrogen n=3→2n{=}3\rightarrow2 Balmer transition at 656.3 nm. Because the n=2n{=}2 level population of quiet H I gas is negligible, Hα is effectively unabsorbed and traces the energy-injection site, while conventional Hα backgrounds from star formation and shocks are strongly suppressed in quiescent, metal-poor dwarfs. This is complementary to earlier limits on DM heating of the same gas (Wadekar & Wang, Phys. Rev. D 106, 075007 (2022), arXiv:2111.08025), which constrain the gas temperature rather than recombination emission.

Leo T, a gas-rich transition dwarf about 420 kpc away, is the ideal first target: its H I distribution and mass profile are well constrained, its star formation is minimal, and the Multi Unit Spectroscopic Explorer (MUSE), the integral-field spectrograph on the Very Large Telescope, has observed it deeply without detecting extended Hα (Vaz et al., A&A 678, A59 (2023), the MUSE-Faint survey; an integral-field spectrograph records a spectrum for each spatial pixel, or spaxel). From the reported per-spaxel noise (σF,pix=0.99×10−20 erg s−1 cm−2\sigma_{F,\rm pix}=0.99\times10^{-20}\ \mathrm{erg\,s^{-1}\,cm^{-2}} per 0.2′′×0.2′′0.2''\times0.2'' spaxel), Leane combined all 9×1049\times10^4 spaxels over the 3600 arcsec23600\ \mathrm{arcsec^2} aperture with uniform N\sqrt{N} statistics plus an empirical pixel-covariance penalty, obtaining a one-sided 95% confidence-level (C.L.) Hα flux limit FHαlim=7.8×10−18 erg s−1 cm−2F^{\rm lim}_{\rm H\alpha}=7.8\times10^{-18}\ \mathrm{erg\,s^{-1}\,cm^{-2}}, and translated it into the first Hα-based limits on ⟨σv⟩\langle\sigma v\rangle and τχ\tau_\chi versus DM mass mχm_\chi for photon and electron final states — leading over parts of the eV–GeV range.

A matched filter is the standard optimal-extraction technique for a known signal template in approximately Gaussian noise: instead of weighting all pixels equally, each pixel is weighted by its expected signal divided by its noise (including pixel-to-pixel covariances), which maximizes the signal-to-noise ratio for the assumed spatial shape. Because annihilation traces ρDM2\rho_{\rm DM}^2 (centrally concentrated) while decay traces ρDM\rho_{\rm DM} (more extended), and the DM halo density profile (cusped Navarro–Fren–White (NFW) versus cored Burkert family) differs from the flat aperture average and from the gas distribution, the published uniform-averaging limits are expected to be suboptimal — but the size of the improvement is unknown and depends on the DM profile, the neutral-gas distribution, electron transport into and out of the aperture, and the measured noise covariance.

This is exactly where the source stops and defers work. It states that a matched filter weighting each pixel by the expected DM morphology would provide an additional signal-to-noise handle; that adding the blue Balmer line Hβ (the 4→24\rightarrow2 transition at 486.1 nm) alongside Hα, together with the absence of a metal-line forest (metal recombination lines that accompany conventional ionization sources), would help distinguish a DM recombination signal from residual star formation or shocks; and that its Extremely Large Telescope (ELT) projection — which simply rescales the MUSE flux limit by a factor of about 5 to represent HARMONI, the ELT's first-light integral-field spectrograph, assuming a stack of 10 Leo-T-like dwarfs — should be replaced by a forecast accounting for HARMONI's smaller field of view and the spatial morphology of the signal, since for spatially extended emission a smaller field of view changes the mapping between line-flux sensitivity and DM parameter space. It further notes that the Rubin Observatory's Legacy Survey of Space and Time (LSST) and the Dark Energy Spectroscopic Instrument (DESI) should discover many new Milky Way satellites, with Leo-T-like systems detectable out to roughly 5 Mpc (versus about 0.5 Mpc today), enabling stacked analyses of multiple similarly quiet, gas-rich dwarfs.

Problem Statement

Determine how much the sensitivity of the dwarf-galaxy Hα recombination-line search for dark matter annihilation and decay can be improved by optimized analysis, and derive the resulting or projected limits. The three requested components serve this single objective — establishing the realistic optimized reach of the newly opened recombination-line channel — and are scientifically linked: each one reweights or extends the same Leo-T-type signal model, and their gains compound in the final limit curves.

  1. Matched-filter re-analysis: construct the expected DM Hα spatial template for annihilation (ρDM2\rho_{\rm DM}^2 weighting, the JJ-factor morphology) and decay (ρDM\rho_{\rm DM} weighting, the DD-factor morphology) for Leo T, including the halo-density profile family (cusped Navarro–Fren–White (NFW) versus cored Burkert), the H I gas distribution, the relevant electron-transport and Hα-attenuation effects, and the measured pixel-to-pixel noise covariance; build the corresponding matched filter over the MUSE data (or its published per-spaxel noise characterization); re-derive the 95% confidence-level (C.L.) Hα flux limit and the limits on ⟨σv⟩\langle\sigma v\rangle and τχ\tau_\chi versus mχm_\chi for photon and electron final states; and quantify the signal-to-noise gain over the published uniform aperture averaging.

  2. Multi-line discrimination: extend the extraction to the Hβ Balmer line, predict the DM-induced Hα/Hβ intensity ratio for the recombination cascade in the relevant conditions, and formulate the metal-line-forest test that separates a DM recombination signal from residual star formation or shock excitation.

  3. Stacked and next-generation forecasts: forecast the reach of stacking NN similarly quiet, gas-rich dwarfs of the kind discoverable out to roughly 5 Mpc by Rubin/LSST and the Subaru Hyper Suprime-Cam (HSC), and produce a realistic ELT/HARMONI forecast that replaces the uniform factor-of-about-5 rescaling of MUSE sensitivity by accounting for HARMONI's smaller field of view, spaxel scale, and sensitivity mapped onto the spatially extended, centrally weighted DM signal.

Deliverables 1–3 jointly answer: what is the optimized sensitivity of the recombination-line DM search, and how much annihilation/decay parameter space can it exclude or project to exclude?

References

  1. Rebecca K. Leane, Search for Dark Matter Annihilation and Decay with Hα Line Emission, Phys. Rev. Lett. 137, 071002 (2026), DOI 10.1103/x38l-5dqc, arXiv:2512.09019, https://arxiv.org/abs/2512.09019
  2. Daniel Vaz, Jarle Brinchmann, Sebastiaan L. Zoutendijk, Leindert A. Boogaard, Sebastian Kamann, Justin I. Read, Martin M. Roth, Peter M. Weilbacher, Matthias Steinmetz, The MUSE-Faint survey IV. Dissecting Leo T, a gas-rich relic with recent star formation, Astron. Astrophys. 678, A59 (2023), DOI 10.1051/0004-6361/202346453, https://doi.org/10.1051/0004-6361/202346453
  3. Digvijay Wadekar, Zihui Wang, Strong constraints on decay and annihilation of dark matter from heating of gas-rich dwarf galaxies, Phys. Rev. D 106, 075007 (2022), DOI 10.1103/PhysRevD.106.075007, arXiv:2111.08025, https://arxiv.org/abs/2111.08025
  4. Simran Arora, Nandini Das, Sukanta Dutta, Ashok Goyal, Dark Photon - ALP Freeze-in: 511 keV and Hα Constraints, arXiv:2607.29131, https://arxiv.org/abs/2607.29131
  5. Giovani Dalla Valle Garcia, Juan Herrero-García, Joel Jones-Pérez, Javier Silva-Malpartida, Opening the parameter space of sub-GeV inelastic dark matter through parity violation, JHEP 04 (2026) 057, DOI 10.1007/JHEP04(2026)057, arXiv:2512.14828, https://arxiv.org/abs/2512.14828
  6. Carlos Blanco, Rebecca K. Leane, Marianne Moore, Joshua Tong, Search for Dark Matter Induced Airglow in Planetary Atmospheres, Phys. Rev. Lett. 136, 251001 (2026), DOI 10.1103/g53c-cvnh, arXiv:2408.15318, https://arxiv.org/abs/2408.15318