Caroline Heneka

Cross-correlations between 21 cm observations and galaxy surveys provide a powerful probe of reionization by providing robustness against foreground contamination while linking ionization morphology to galaxies. We quantified the constraining power of 21 cm galaxy cross-power spectra for inferring the neutral hydrogen fraction, x _HI ( z ), and mean overdensity, ⟨1 + δ _HI ⟩( z ), exploring dependence on the field of view; redshift precision, σ _z ; and minimum halo mass, M _h, min . We employed our simulation-based inference framework EoRFlow for likelihood-free parameter estimation. Mock observations include thermal noise for 100 h of SKA-Low with foreground avoidance and realistic galaxy-survey effects. For a fiducial survey (FOV = 100 deg ² , σ _z = 0.001, M _h, min = 10 ¹¹ M _⊙ ), cross-power spectra yield unbiased constraints with posterior volumes (PVs) of ∼10% relative to priors. Cross-power measurements reduce the PV by 20–30% versus 21 cm auto-power alone. With foreground avoidance, spectroscopic redshift precision is essential; photometric redshifts render cross-correlations uninformative. Notably, cross-power spectra constrain ionizing source properties, the escape fraction f _esc, and the star formation efficiency f _* , which remain degenerate in auto-power (PV > 60%). Tight constraints require either deep surveys detecting faint galaxies ( M _h, min ∼ 10 ¹⁰ M _⊙ ) with moderate foregrounds (PV ∼ 11%) or conservative mass limits with optimistic foreground removal (PV ∼ 19%). 21 cm galaxy cross-correlations enhance morphology constraints beyond auto-power while enabling previously inaccessible source property constraints. Realizing full potential requires precise redshifts and either faint galaxy detection limits or improved 21 cm foreground cleaning.

Context. The 21cm–galaxy cross-correlations are expected to play a key role in the confirmation of the cosmological 21cm signal. Aims. We investigate which survey configurations detect the 21cm–Lyman- α emitter (LAE) cross-correlation signal and assess its ability to distinguish reionisation scenarios. Methods. Our pipeline computes observational uncertainties for the 21cm–galaxy cross-power spectrum, accounting for key survey parameters. These include field of view (FoV); limiting luminosity of galaxy surveys, L _α ; redshift uncertainty, σ _z ; and 21cm foreground wedge assumptions. Using this pipeline, we calculated the signal-to-noise ratio (S/N) of the 21cm–LAE cross-power spectrum for two scenarios: one where faint galaxies dominate the ionising photon budget and one where reionisation is driven by bright galaxies. Results. We find that: (i) S/N increases with a larger FoV, fainter L _α , and smaller σ _z , with the FoV having the strongest impact when σ _z is small. (ii) Under a moderate foreground wedge, photometric-like surveys yield an insufficient S/N, and medium-deep ( L _α ≳ 10 ^42.5 erg s ⁻¹ ), wide-area (FoV > 20 deg ² ) slitless spectroscopic or spectroscopic surveys are required. (iii) Under an optimistic foreground wedge, detection is possible with deep ( L _α ≳ 10 ^42.3 erg s ⁻¹ ), wide-area (FoV ≳ 80 deg ² ) photometric-like surveys, or with shallower, small-area (FoV ≃ 2 − 3 deg ² ) slitless spectroscopic surveys. (iv) To distinguish the two reionisation scenarios at z = 7, moderate foreground wedge scenarios require deep-wide spectroscopic surveys; under optimistic foreground wedge, assumptions shallower ( L _α ≃ 10 ^42.8 erg s ⁻¹ ), medium-area (FoV ≃ 10 deg ² ) slitless spectroscopic surveys suffice. (v) Maximising the S/N for both detection and model discrimination requires sampling the large-scale peak of the cross-power spectrum, which shifts to larger physical scales as reionisation proceeds, and the less ionisation fronts follow the gas density, thus making surveys at z > 7, or when the Universe was more neutral (⟨ χ _HII ⟩> 0.5), more promising despite lower galaxy number densities. Conclusions. Our results show that large-area spectroscopic surveys and 21cm foreground cleaning are key for using 21cm–LAE cross-correlations to constrain reionisation beyond the global ionisation state of the intergalactic medium.

Modern machine learning techniques can unlock the vast cosmological information encoded in forthcoming Square Kilometre Array (SKA) observations. We show that tomographic 21 cm data from the reionisation era can yield stringent tests of inflationary models — here illustrated with Starobinsky R + R ² inflation. Using a simulation-based inference (SBI) framework, we compare neural summaries (convolutional network and vision transformer) with a traditional power spectrum summary and perform a fully joint SBI analysis combining 21 cm data with data of the cosmic microwave background (CMB). Forecasts based on realistic mock observations indicate that SKA alone will achieve constraints competitive with Planck, and that the combined SKA + CMB dataset will tighten bounds on both inflationary and ΛCDM parameters considerably while improving precision on key astrophysical quantities.

The 21cm line from the spin-flip transition of neutral hydrogen (HI) provides a unique window into the Epoch of Reionization (EoR), the final phase transition of our Universe. The Square Kilometre Array (SKA) enables precise measurements of 21cm fluctuations that trace ionization, temperature, and density fluctuations of the intergalactic medium (IGM). Nevertheless, a direct reconstruction of the timeline of the EoR in terms of the progress of ionization remains an ongoing challenge due to the highly non-Gaussian nature and thus intractable likelihood of the 21cm signal. Here, we present EoRFlow , a simulation-based inference (SBI) framework for reconstructing the global neutral hydrogen fraction x _HI ( z ) directly from 2D cylindrically averaged power spectra (2DPS) of the 21cm signal. We validate our method on realistic mock datasets for SKA-Low. Bypassing the need for explicit likelihood formulations, our approach enables fast, unbiased posterior estimation of the x _HI evolution in narrow redshift slices, allowing for piecewise reconstruction of the global reionization history. By directly inferring the reionization history from 21cm power spectra, our framework provides a scalable and robust path forward for 21cm cosmology in the SKA era.

Context. The second data release of the LOFAR Two-Metre Sky Survey (LoTSS-DR2) extends the first data release in terms of sky coverage and source density. It provides the largest radio source catalogue to date, including 4.4 million sources and covering 5635 square degrees of the sky. Therefore, it provides an excellent opportunity for studies of the large-scale structure of the Universe. Aims. We investigated the statistical distribution of source counts in cells and we tested a computationally cheap method based on the counts in cells to estimate the two-point correlation function. Methods. We studied and compared three stochastic models for the counts in cells; these resulted in a Poisson distribution, a compound Poisson distribution, and a negative binomial distribution. By analysing the variance of counts in cells for various cell sizes, we fitted the reduced normalised variance to a single power-law model representing the angular two-point correlation function. Results. Our analysis confirms that radio sources are not Poisson distributed, which is most likely due to multiple physical components of radio sources. Employing instead a Cox process, we show that there is strong evidence in favour of the negative binomial distribution above a flux-density threshold of 2 mJy. Additionally, the mean number of radio components derived from the negative binomial distribution is in good agreement with corresponding estimates based on the value-added catalogue of LoTSS-DR2. The scaling of the counts-in-cells normalised variance with cell size is in good agreement with a power-law model for the angular two-point correlation. At a flux-density threshold of 2 mJy and a signal-to-noise ratio of 7.5 for individual radio sources, we find that for a range of angular scales large enough to not be affected by the multi-component nature of radio sources, the value of the exponent of the power law ranges from −0.8 to −1.05. This closely aligns with findings from previous optical, infrared, and radio surveys of the large-scale structure. Conclusions. The multi-component nature of LoTSS radio sources is essential in order to understand the observed counts-in-cells statistics. The scaling of the counts-in-cells statistics with cell size provides a computationally efficient method for estimating the two-point correlation properties, offering a valuable tool for future large-scale structure studies.

Aims. We cross-correlated galaxies from the LOw-Frequency ARray (LOFAR) Two-metre Sky Survey (LoTSS) second data release (DR2) radio source with the extended Baryon Oscillation Spectroscopic Survey (eBOSS) luminous red galaxy (LRG) sample to extract the baryon acoustic oscillation (BAO) signal and constrain the linear clustering bias of radio sources in LoTSS DR2. Methods. In the LoTSS DR2 catalogue, employing a flux density limit of 1.5 mJy at the central LoTSS frequency of 144 MHz and a signal-to-noise ratio (S/N) of 7.5, additionally considering eBOSS LRGs with redshifts between 0.6 and 1, we measured both the angular LoTSS-eBOSS cross-power spectrum and the angular eBOSS auto-power spectrum. These measurements were performed across various eBOSS redshift tomographic bins with a width of Δz = 0.06. By marginalising over the broadband shape of the angular power spectra, we searched for a BAO signal in cross-correlation with radio galaxies, and determine the linear clustering bias of LoTSS radio sources for a constant-bias and an evolving-bias model. Results. Using the cross-correlation, we measured the isotropic BAO dilation parameter as α = 1.01 ± 0.11 at z_eff = 0.63. By combining four redshift slices at z_eff = 0.63, 0.69, 0.75, and 0.81, we determined a more constrained value of α = 0.968_−0.095^+0.060. For the entire redshift range of z_eff = 0.715, we measured b_C = 2.64 ± 0.20 for the constant-bias model, b(z0) = b_C, and then b_D = 1.80 ± 0.13 for the evolving-bias model, b(z) = b_D/D(z), with D(z) denoting the growth rate of linear structures. Additionally, we measured the clustering bias for individual redshift bins. Conclusions. We detected the cross-correlation of LoTSS radio sources and eBOSS LRGs at a 9.2σ statistical significance for one single redshift bin and at a 14.7σ significance when the four redshift bins were combined. For the BAO signal, we achieved a significance of 2.2σ for a single redshift bin, 2.7σ for the combined cross-correlation and eBOSS auto-correlation, and 4σ for the combined analysis of four redshift bins in the cross-correlation, when assuming a Gaussian distribution for the BAO dilation parameter.

The size and complexity reached by the large sky spectroscopic surveys require efficient, accurate, and flexible automated tools for data analysis and science exploitation. We present the Galaxy Spectra Network/GaSNet-II, a supervised multinetwork deep learning tool for spectra classification and redshift prediction. GaSNet-II can be trained to identify a customized number of classes and optimize the redshift predictions. Redshift errors are determined via an ensemble/pseudo-Monte Carlo test obtained by randomizing the weights of the network-of-networks structure. As a demonstration of the capability of GaSNet-II, we use 260k Sloan Digital Sky Survey spectra from Data Release 16, separated into 13 classes including 140k galactic, and 120k extragalactic objects. GaSNet-II achieves 92.4 per cent average classification accuracy over the 13 classes and mean redshift errors of approximately 0.23 per cent for galaxies and 2.1 per cent for quasars. We further train/test the pipeline on a sample of 200k 4MOST (4-metre Multi-Object Spectroscopic Telescope) mock spectra and 21k publicly released DESI (Dark Energy Spectroscopic Instrument) spectra. On 4MOST mock data, we reach 93.4 per cent accuracy in 10-class classification and mean redshift error of 0.55 per cent for galaxies and 0.3 per cent for active galactic nuclei. On DESI data, we reach 96 per cent accuracy in (star/galaxy/quasar only) classification and mean redshift error of 2.8 per cent for galaxies and 4.8 per cent for quasars, despite the small sample size available. GaSNet-II can process ∼40k spectra in less than one minute, on a normal Desktop GPU. This makes the pipeline particularly suitable for real-time analyses and feedback loops for optimization of Stage-IV survey observations.