Mapping neutral islands during end stages of reionization with photometric intergalactic medium tomography

TL;DR

This work investigates mapping remaining neutral islands at the end of reionization (z ≲ 6) using photometric IGM tomography with deep NB imaging (NB816). It combines large-scale radiative-transfer–powered reionization simulations with Lyα transmission modeling to forecast the observability of neutral islands at z ≈ 5.7 and to predict cross-correlations with galaxies, 21-cm signals, and the patchy CMB optical depth. A novel pattern-recognition pipeline using superpixels and thresholding enables identification of large neutral islands in noisy tomographic maps, recovering island-size distributions to ~0.3 dex. The study demonstrates that, with sufficient background sources (≳500 per (300 Mpc)^2) and NB depth (≈27.5 mag at 3σ), photometric IGM tomography can reveal neutral islands and provide cross-checks via multiple observables, informing the morphology and timeline of cosmic reionization and guiding future follow-up with facilities like SKA, CMB-S4, and upcoming galaxy surveys.

Abstract

During the epoch of reionization (EoR), the first generation of luminous sources in our Universe emitted ionizing photons that almost completely ionized the gas in the intergalactic medium (IGM). The growth of ionized bubbles and the persistence of neutral islands within the IGM hold vital clues to understanding the morphology and timeline of cosmic reionization. We explore the potential of photometric IGM tomography using deep narrow-band (NB) imaging to observe the Lyman-$α$ forest transmission in background galaxies with the Subaru/Hyper-Suprime Cam (HSC). Based on our simulations, we find that the currently available NB filter is suitable for mapping the IGM at $z\simeq 5.7$, corresponding to the late stages of reionization. Our findings indicate that over $\sim$500 background galaxies are needed to accurately reconstruct the IGM at scales greater than 200 Mpc, achieving more than a 40 per cent correlation with the true distribution. This technique can help detect final remaining neutral islands that span more than 20 Mpc lengths. Using the superpixel method built to identify physical patterns in noisy image data, we find that the neutral island size distribution can be recovered with an accuracy of $\sim$0.3 dex. Furthermore, we demonstrate that these reconstructed maps are correlated with the galaxy distribution and anti-correlated with the cosmological 21-cm signal from neutral hydrogen in the IGM. Lastly, we find that these reconstructed maps are anti-correlated with the patchy optical depth to the cosmic microwave background. As such, multiple measurements can be employed for confirmed detection of neutral islands during the end stages.

Figures (13)

• Figure 1: Slices from the two reionization models at redshift $z=5.7$. Top panels: We show the matter density contrast slice on the left. In the Uniform_Reionization model, the universe has a uniform neutral fraction of $x_\mathrm{HI}=10^{-4}$ everywhere. On the right, we show the map of the effective optical depth of Lyman-$\alpha$ photons $\tau_\mathrm{eff}$ at simulation resolution in the field of view direction. These values are averaged over 45 Mpc length scales along the line-of-sight. The black circle gives the field of view of the Subaru/HSC NB816 filter. Bottom panels: We present a neutral fraction slice and the corresponding $\tau_\mathrm{eff}$ in the left and right, respectively, panels for the Late_Reionization model. In all the Lyman-$\alpha$ effective optical depth maps, we can see that the low values of transmissions correspond to the neutral hydrogen distribution.
• Figure 2: The surface number density of background galaxies for photometric IGM tomography at $z=5.7$ as a function of the apparent limiting UV background of the background source population. The various curves indicate the spectroscopic confirmation rates of $\epsilon_{\rm spec}=0.8$ (dashed), $0.2$ (solid), and $0.1$ (dotted). The right y-axis shows the corresponding typical angular separation between background galaxies, which determines the spatial resolution of the tomographic map.
• Figure 3: The number of background galaxies required for IGM tomography. Top-left panel: The Spearman correlation coefficient of the ground truth with the reconstructed transmission maps applying different brightness ($m_\mathrm{UV}$) cuts on the luminosity function. The x-axis on the top gives the number of background sources corresponding to these cuts. We see that the correlation improves with an increase in the number of background sources. Top-right to bottom-left in clockwise: The reconstructed Lyman-$\alpha$ effective optical depth maps employing brightness limits at $m_\mathrm{UV}=26$, 25.5 and 25, respectively. The black circles represent the positions of the background sources. The resolution of the maps decreases with the number of background sources, which causes a decrease in correlation with the ground truth.
• Figure 4: The normalized histogram of the Lyman-$\alpha$ forest transmission $\mathcal{T}$, which is shown in the right panels of Fig. \ref{['fig:models_tau_map']}. The mean value of the map $\langle \mathcal{T}\rangle$ is subtracted before estimating these histogram to focus on their shape. The solid curves represent the ground truth of the two reionization models, while the dashed lines show the histogram of the reconstructed transmission maps assuming background galaxies down to $m_{\rm UV}^{\rm limit}=26.0$. The peak of the histogram for the simulation with neutral islands ( Late_Reionization) shifts towards lower transmission values. The reconstructed Lyman-$\alpha$ transmission maps exhibit more skewness for the Late_Reionization compared to Uniform_Reionization model.
• Figure 5: Feasibility of recovering coherent IGM features from photometric IGM tomographic maps. We assume a $3\sigma$ narrow-band (NB) depth of $n_{\rm NB}^{\rm limit} = 27.5$. The limiting UV magnitudes are $m_{\rm UV}^{\rm limit} = 26.0$ (left panels) and $25.5$ (right panels). The top, middle, and bottom rows show the noisy reconstructed IGM tomographic maps, the corresponding standard deviation maps, and the signal-to-noise ratio (SNR) maps, respectively. The contours of the neutral islands at simulation resolution is shown in the top panel and simulated positions of background galaxy sight-lines are indicated by black circles in the bottom two panels.