Dark Ages 21 cm Signal

• Dark Ages 21 cm Signal is the redshifted hyperfine transition of neutral hydrogen that probes the pre-luminous Universe and its thermal state.
• Bayesian methods and model comparisons quantify spectral features, with absorption troughs reaching depths up to -161 mK across 1–50 MHz.
• Observational strategies demand ultra-sensitive, wide-band measurements to overcome dominant Galactic foregrounds, favoring lunar-farside or space-based experiments.

The Dark Ages 21 cm signal refers to the redshifted hyperfine transition of neutral hydrogen (HI), observed at frequencies ν ≈ 50–1 MHz (corresponding to redshifts z ≈ 30–1100), which directly probes the physical and cosmological state of the Universe prior to the emergence of the first luminous sources. This epoch, known as the "Dark Ages," is uniquely accessible through the 21 cm line, providing an unrivaled window into cosmological initial conditions, the thermal and ionization history of the intergalactic medium (IGM), and possible exotic physics. Theoretical predictions, detectability forecasts, and discriminability between cosmological models are now well quantified using Bayesian methods and physically motivated foreground assumptions (Yoshiura et al., 1 Feb 2026). The global signal, power spectrum, and higher-order statistics (e.g., trispectrum) each encode distinct aspects of matter and radiation in the high-redshift Universe.

1. Theoretical Basis for the Dark Ages 21 cm Signal

The sky-averaged (global) 21 cm brightness temperature relative to the radio background $T_R$ is governed by collisional and radiative processes that set the spin temperature $T_S$ of neutral hydrogen. The differential brightness temperature as a function of redshift $z$ (or frequency $\nu=1420\,\mathrm{MHz}/(1+z)$) is

$$\delta T_b(z) \simeq 27\, x_{\rm HI} (1 + \delta_b) \left( \frac{\Omega_b h^2}{0.023} \right) \left( \frac{0.15}{\Omega_m h^2} \frac{1+z}{10} \right)^{1/2} \left[1 - \frac{T_R(z)}{T_S(z)}\right] \,\mathrm{mK}$$

where:

• $x_{\rm HI}$: neutral hydrogen fraction ($\sim$1 throughout the Dark Ages),
• $\delta_b$: baryon overdensity (zero for the sky-average),
• $$T_R(z) = T_{\rm CMB}(z) + T_{\rm ERB}(\nu/\nu_{78})^{-2.6}$$: total radio background; in standard cosmology, $T_{\rm ERB}=0$,
• $T_S(z)$: spin temperature, set by competition between radiative ($T_{\rm CMB}$) and collisional (gas kinetic temperature, $T_K$) bath,
• $\Omega_b h^2, \Omega_m h^2$: baryonic and total matter physical densities.

During $30 \lesssim z \lesssim 200$, collisional coupling ($x_c \gg 1$) ensures $T_S \approx T_K \ll T_{\rm CMB}$, resulting in a broad absorption trough ($\delta T_b < 0$) with a minimum at $\nu \sim 16$ MHz and depth $\sim -40$ mK in the standard $\Lambda$CDM model. At higher $z$, $T_K$ and $T_S$ are locked to the CMB by Compton scattering; at lower $z$, $x_c \ll 1$ and $T_S \to T_{\rm CMB}$, quenching the signal (Yoshiura et al., 1 Feb 2026, Bevins, 12 Dec 2025).

2. Predicted Spectral Features and Model Space

Eight physically distinct, non-astrophysical cosmological models have been computed (by modifying RECFAST), each yielding a characteristic absorption or emission morphology in the 1–50 MHz band. Key features include:

Model	$\Delta T_b^{\rm min}$	Peak Frequency	Process
$\Lambda$CDM	$-40$ mK	$16$ MHz	Standard, adiabatic cooling
DMBw	$-62$ mK	$17$ MHz	Weak DM-baryon cooling
DMBs	$-161$ mK	$14$ MHz	Strong DM-baryon cooling
EDE	$-58$ mK	$14$ MHz	Early dark energy
ERB	No clear trough	–	Excess radio background
LDMD	$-30$ mK	$16$ MHz	Light DM decay heating
PMFw	$-21$ mK	$17$ MHz	Weak primordial $B$-field heating
PMFs	$+26$ mK (emission)	$9$ MHz	Strong primordial $B$-field heating

Models with slow, smooth frequency evolution (ERB, PMFw, LDMD) are subject to strong foreground degeneracies (Yoshiura et al., 1 Feb 2026). Distinctive, sharp features (DMBw, DMBs) are most robustly discriminable (Paul et al., 24 Oct 2025, Novosyadlyj et al., 2024). The feature positions and depths are exceptionally stable under $\Lambda$CDM parameter uncertainties: for Planck constraints, depth is known to within $<\!1$ mK and frequency to $<\!0.05$ MHz (Bevins, 12 Dec 2025).

3. Foregrounds and Degeneracies

The dominant foreground at $\nu < 50$ MHz is Galactic synchrotron emission, modeled as: $$T_{\rm FG}(\nu) = T_G(\nu/10\,{\rm MHz})^{\alpha-2} [1 - e^{-\tau(\nu)}]/\tau(\nu) + T_{\rm ex}(\nu/10\,{\rm MHz})^{\beta-2} e^{-\tau(\nu)}$$ where $\tau(\nu)=F\nu^{-2.1}$. With $T_G \sim 2.5 \times 10^{5}$ K and $T_{\rm ex} \sim 5.5 \times 10^4$ K, the total foreground reaches $\sim 10^7$ K at 1 MHz, vastly exceeding the cosmological signal [$|\delta T_b| \lesssim 0.1$ K]. The smoothness of $T_{\rm FG}(\nu)$ means that only 21 cm features with sufficient spectral structure can be separated; models with long-wavelength variation are subject to partial absorption into foreground parametrization (Yoshiura et al., 1 Feb 2026, Burns et al., 2021). Foreground modeling is further complicated by spatial anisotropy, ionospheric refraction (for ground arrays), polarization leakage, and instrumental systematics, none of which substantively alter the foreground's spectral smoothness but add complexity to separation strategies.

4. Experimental Strategies and Sensitivity Requirements

Observational configurations studied include:

• Wide-band, continuous coverage (A): 1–50 MHz, 1 MHz channels, 10,000 h integration, per-channel noise floor $\sim$5 mK. This allows robust detection ($>5\sigma$) of the standard trough and discrimination among deeply absorbing models.
• Sparse, multi-channel “narrow-band” (B): 11 bands at 1,5,10,…,50 MHz. At 5 mK/channel, only strongest-trough models are distinguishable; improved sensitivity (1 mK) restores wide-band performance.
• Limited coverage (C): 15–50 MHz. Excluding 1–15 MHz leaves key features degenerate with foregrounds, precluding detection.

Quantitatively, with configuration (A), standard $\Lambda$CDM yields $\Delta\ln Z>5$ compared to the null-signal: detection is “very strong” in the Bayesian model selection sense. Sharp-trough variants (DMBs) can reach $\Delta\ln Z\sim70$—evidence is overwhelming. Models with flat troughs or weak features (LDMD, PMFw, ERB) can only be detected if calibration and per-channel sensitivity are pushed to $\lesssim$1 mK. Low-frequency coverage below 15 MHz is essential to break foreground degeneracies (Yoshiura et al., 1 Feb 2026). Terrestrial ionospheric opacity prevents ground-based surveys below $\sim$15 MHz; the full 1–50 MHz band is only accessible via lunar-farside or space-based instruments (Burns et al., 2021, Burns et al., 2021).

5. Bayesian Evidence and Model Discrimination

Detection and model-selectivity analyses employ the integrated Bayesian evidence,

$$Z_i = \int d\theta\, \mathcal{L}(D|\theta, M_i) \pi(\theta|M_i)$$

with model comparison via $\Delta\ln Z_{i,j} = \ln Z_i - \ln Z_j$, with standard interpretive thresholds (Kass & Raftery 1995): $<1$ not worth mentioning; $1-3$ positive; $3-5$ strong; $>5$ very strong. Distinctive, deep absorption features (DMBw, DMBs) exhibit $\Delta\ln Z\gg5$ relative to all alternatives under realistic sensitivities. $\Lambda$CDM vs. EDE is marginal, with $\Delta\ln Z\sim1$–$2$ (positive but not decisive), reflecting closely matched troughs. Smooth-spectrum models and those with low-frequency peaks/plateaus are degenerate with spectrally adjustable foregrounds unless channel sensitivity is enhanced below 1 mK or ultra-wide bandwidth is used (Yoshiura et al., 1 Feb 2026). Systematic effects—chromatic gain, polarization leakage, calibration errors—were not included, and in practice set a demanding floor for detection.

6. Cosmological and Fundamental Physics Implications

Detection of a “standard” $\Lambda$CDM trough (|$\delta T_b| \sim 40$ mK at $\nu_p\sim16$ MHz) affirms the paradigm of adiabatic baryonic cooling and neutral hydrogen dominance in the pre-luminous Universe (Yoshiura et al., 1 Feb 2026, Bevins, 12 Dec 2025). Measurements of trough depth and spectral shape provide:

• Joint constraints on small-scale matter power (e.g., suppression by warm/fuzzy DM, potential deviation from scale invariance) (Park et al., 14 Sep 2025, Novosyadlyj et al., 2024).
• Limits on baryon–dark-matter interaction (e.g., Rutherford/coulombic cooling, co-SIMP, annihilating/decaying DM). Current best bounds are $$f_\chi^2 \langle \sigma v \rangle/M_\chi \lesssim 10^{-28}$$ cm$^3$ s$^{-1}$ GeV$^{-1}$, nearly an order of magnitude tighter than CMB limits (Mohapatra, 25 Jun 2025, Paul et al., 24 Oct 2025).
• Exclusion of models with nonstandard radio backgrounds (e.g., from decaying particles or primordial magnetic fields) if no deviations from the standard trough are observed (Mondal et al., 2023).
• Direct constraints on early dark energy; strong distinguishability only in the case of non-smooth, deep features.
• Sensitivity to primordial non-Gaussianity and inflationary signatures through higher-order moments in forthcoming fluctuation measurements (Flöss et al., 2022).

Absence of a trough, or unambiguous detection of emission or a shallower/minimum feature, would require a substantial revision of energy-injection and cooling physics in the early Universe.

7. Experimental Limitations, Outlook, and Future Prospects

Key technical challenges include:

• Achieving $\lesssim$5 mK noise per MHz across 10,000 h—demanding exceptional receiver stability and calibration linearity.
• Mitigating smooth, high-dynamic-range foregrounds whose amplitudes exceed the cosmological signal by $>10^6$. Only features with sufficient spectral structure (width $\ll$width of foreground fit, central frequency coverage) are guaranteed to be distinguishable (Yoshiura et al., 1 Feb 2026, Pober et al., 30 Jul 2025).
• Ground-based arrays are limited to $\nu \gtrsim 15$ MHz by the ionosphere; lunar-farside or space-based missions are mandatory for the full dark-ages band (Burns et al., 2021, Burns et al., 2021).
• Ignoring systematics (beam chromaticity, calibration) and cosmic-ray transients is a theoretical abstraction; practical implementation must confront these at the sub-mK level.

Nevertheless, under physically realistic, idealized conditions, a wide-band measurement (1–50 MHz, $\lesssim 5$ mK sensitivity, 10,000 h) is predicted to definitively detect the standard dark-ages trough at $>5\sigma$, and discriminate among models with high confidence—provided the experiment includes the low-frequency end of the band (Yoshiura et al., 1 Feb 2026). Strategies employing sparse frequency coverage or limited channel counts require severalfold better per-channel sensitivity or recourse to distinctive, extrinsic spectral features.

A successful detection of the Dark Ages 21 cm global signal will furnish a fundamentally new test of cosmology and of possible new physics, unencumbered by astrophysical uncertainties. It is a principal science goal for planned lunar-farside and deep-space radio observatories.