21 cm Cosmology Sensitivity to Small-Scale Structure: Warm vs Neutrino-Interacting Dark Matter

TL;DR

The paper addresses whether near-future 21 cm observations can differentiate two non-cold dark matter scenarios—thermal warm dark matter (WDM) and neutrino–dark matter interactions (νDM)—by connecting them through a common cutoff scale $\lambda_{\rm cut}$ and forecasting HERA’s capabilities with Fisher analyses. It develops a unified transfer-function framework $T_i(k)=\big[1+(\lambda_{\rm cut} k)^{\gamma}\big]^{-\delta}$, calibrated to CLASS results, and links $\lambda_{\rm cut}$ to the WDM mass $m_{\rm WDM}$ and νDM coupling $u_{\nu\rm DM}$, enabling consistent cross-model comparisons. Using 21cmFAST with two galaxy populations (ACGs and MCGs) and a 1000-hour HERA forecast, the work reveals detectable νDM interaction strengths down to $\sigma_{\nu\rm DM}$ of about $3\times 10^{-35}$ cm$^{2}$ for GeV DM, but finds that HERA cannot decisively distinguish νDM from WDM due to degeneracies with astrophysical parameters. When a νDM signal is assumed, the corresponding WDM mass would be up to $\sim 9$ keV for optimistic modelling, implying that 21 cm data could test or challenge related Lyman-$\alpha$ bounds and recent νDM hints. Overall, the paper highlights that disentangling DM microphysics from early-universe astrophysics hinges on reducing modelling uncertainties, and it clarifies the practical prospects and limitations of using the 21 cm signal to differentiate these NCDM scenarios.

Abstract

The $21\,$cm signal originating from Cosmic Dawn to the Epoch of Reionisation is highly sensitive to the processes governing star formation in the early universe as well as new physics. In this work, we focus on the imprint of non-cold dark matter (DM), which impacts the formation of the smallest halos. Our goal in particular is to clarify whether near-future radio telescopes such as the Hydrogen Epoch of Reionisation Array (HERA), will be able to distinguish between free-streaming dark matter, specifically in the form of thermal warm DM (WDM), and collisional damping due to neutrino-DM ($ν$DM) interactions giving rise to larger overdensities on small scales. For that purpose we first implement a mapping between the two models in terms of a cutoff scale and determine detection thresholds for the two DM models. Using Fisher matrix forecasts, we show that $ν$DM interaction strengths down to $σ_{ν{\rm DM}}\sim 3\times10^{-35}$ cm$^2$ could be probed by $21\,$cm cosmology when considering two populations of galaxies for a GeV mass DM. This would allow to either confirm or rule out a recent claimed preference for a non-zero $ν$DM interaction in Lyman-$α$ data. Furthermore, we find that HERA will not be able to distinguish between $ν$DM and WDM. In the latter context, the threshold for detection of $ν$DM interactions translates into WDM with mass up to $m_{\rm WDM}\sim 9$ keV that could be detected by HERA.

Figures (13)

• Figure 1: Left: The matter power spectrum at $z=0$ as a function of the comoving wavenumber. The $\Lambda{\rm CDM}+3 m_\nu$ scenario is shown as the black solid line for reference. The matter power spectra of each of the three scenarios that we consider, WDM, ${\nu{\rm DM}}$ and the fitted transfer function, are shown as dotted, dashed and solid lines respectively for two values of the cutoff scale $\lambda_{\rm cut}=2\times10^{-3}\,{\rm Mpc}/h$ (red) and $\lambda_{\rm cut}=4\times10^{-3}\,{\rm Mpc}/h$ (yellow). The corresponding $m_{\rm WDM}$ and $u_{\nu{\rm DM}}$ are shown in Table \ref{['tab:lcut_conversion']}. Right: The square root of the ratios between the matter power spectra of the three scenarios shown in the left plot and the $\Lambda{\rm CDM}+3m_\nu$ matter power spectrum. This corresponds to the transfer functions as defined in Eq. (\ref{['eq:transfer_funtion_definition']}) for $i=$ WDM (dotted) and ${\nu{\rm DM}}$ (dashed) as well as the analytical expression for the transfer function given in Eq. (\ref{['eq:transf_function_fit_form']}) (solid).
• Figure 2: Upper panel Stellar mass distribution in ACGs and MCGs assuming CDM (black) and for three cutoff scales (coloured) considering the reference astrophysical model of Tab. \ref{['tab:fiducial_astro_parameters']} and using the fitting function of Eq. (\ref{['eq:transf_function_fit_form']}) to obtain the NCDM matter power spectrum. The different line styles show the respective contributions from ACGs and MCGs. The vertical lines in the upper panel indicate the values of $M_{\rm turn}^{\rm II}$ and $M_{\rm turn}^{\rm III}$. Lower panel Corresponding halo mass function for CDM and the three values of $\lambda_{\rm cut }$ considered in the upper panel. The dash-dotted vertical lines show the halo mass $M^{\rm SK}_{1/2}$ corresponding to a wave number at which the transfer function is equal to $1/2$, see text for details.
• Figure 3: NCDM imprint on the $21\,$cm signal as a function of the redshift using the fitting function as an input. The upper panel displays the power spectrum at wavelength $k=0.20\;h/\rm Mpc$, the lower panel shows the global signal. The black line assumes CDM and the coloured lines show the effect of different values of $\lambda_{\rm cut}$. The dark and light shaded bands respectively indicate the associated 1 and 2$\sigma$ measurement noise of HERA, assuming 1000 hours of observation and a 20% modelling error (see also section \ref{['sec:analysis']}). We assume our reference astrophysical model.
• Figure 4: Same as Fig. \ref{['fig:astro-impact-PS-1']}, but assuming CDM and showing the effect of varying $\alpha_\star^{\rm III}$ (left panel) and $\alpha_{\rm esc}$ (right panel).
• Figure 5: Marginalised $68\%$ (pale shaded area) and $95\%$ (dark shaded area) confidence intervals of $\lambda_{\rm cut}$, normalised to the value of $\lambda_{\rm cut}^{\rm fid}$, as a function of $\lambda_{\rm cut}^{\rm fid}$. The results are obtained by applying the fitted transfer function of Eq. (\ref{['eq:transf_function_fit_form']}) to the CDM matter power spectrum (input method \ref{['it:fit']}). A conversion between $\lambda_{\rm cut}$, $m_{\rm WDM}$ and $u_{\nu{\rm DM}}$ is also provided. The left panel corresponds to $20\%$ modelling noise ($\varepsilon=0.2$) while the right panel assumes no modelling noise ($\varepsilon=0$). The markers show the obtained marginalised posteriors for each value of $\lambda_{\rm cut}^{\rm fid}$ for which we have effectively performed a Fisher forecast, and the corresponding error bars, see text for details.