The THESAN project: Lyman-alpha intensity mapping of cosmic reionization

TL;DR

This paper develops a pipeline to predict Lyman-α LIM signals during the Epoch of Reionization using the THESAN radiation-hydrodynamic simulations, enabling high-resolution light-cone predictions that incorporate absorption, emission, and damping-wing effects.By rendering Lyα fields on a redshift-space grid and performing LoS radiative transfer with Voigt profiles, the authors quantify contributions from recombination, collisional excitation, and unresolved HII regions, and explore the impact of outflows via a simple velocity-offset model.Key findings include that absorption-included Lyα fluctuations are several orders of magnitude fainter than emission-only predictions, that redward outflows can boost power by up to ~4× at small scales, and that the small-scale power slope steepens during reionization; comparisons with SPHEREx sensitivities demonstrate potential detectability under certain modeling choices.These results underscore the importance of incorporating resonant scattering in IGM radiative transfer and outline next steps, such as including scattering effects, interloper treatment, and larger-volume simulations to extend LIM forecasts to the relevant angular scales.

Abstract

Line Intensity Mapping (LIM) has garnered attention as a powerful cosmological probe, with next-generation instruments such as SPHEREx preparing to map the evolution of large-scale structure during the Epoch of Reionization (EoR). Lyman-alpha emission in the EoR is strongly shaped by resonant absorption from neutral hydrogen in the diffuse intergalactic medium (IGM), which transforms galactic sources into a low surface-brightness background. In this work, we leverage the state-of-the-art THESAN cosmological simulations to produce high-resolution theoretical predictions for future Lyman-alpha LIM studies, constructing continuous light cones for line-of-sight cosmological integrations. We assess the contributions of recombination, collisional excitation, and unresolved HII regions to the total Lyman-alpha spectral intensity. In addition, we explore the IGM in absorption at different redshifts using damping wing analysis. We produce channel maps exploring spatial fluctuations across redshift bands probe-able by LIM instruments. We find that the slope of the absorption-included Lyman-alpha fluctuation power spectrum at smaller scales (k > 10^(-2) 1/arcsec) steepens toward lower redshift, and that our emission-only Lyman-alpha power spectrum lies above the SPHEREx sensitivity, whereas the absorption-included signal is ~4 orders of magnitude lower--providing a conservative lower limit on inhomogeneity signatures and highlighting the importance of including resonant scattering in our model in the future. We also find that including outflows in a simple toy model boosts power by four orders of magnitude. We identify limitations in our analysis and propose next steps, including incorporating the effects of resonant Lyman-alpha scattering and line interlopers, as well as larger simulation volumes.

Figures (23)

• Figure 1: A schematic illustrating the process of stitching renders together to form a continuous light slab. Step 1: The leftmost box shows the first render as a 3D cube, and the shaded pink area corresponds to the data included in our stitched product, sliced along the third (LoS) dimension according to the allocated comoving distance range $\Delta L_0$. Step 2: For the second render, we continue from the point we had stopped at in the first render to ensure continuity along the LoS axis. Step 3: If $\Delta L_1$ exceeds the remaining length in the third dimension, we wrap around the box in the second render to cover the appropriate redshift/comoving distance range. This process is repeated for all subsequent cartesian renders until the desired redshift range is fully covered.
• Figure 2: Ly$\alpha$ damping wing profiles, both along a random sightline (top panel) and when taking the median and $1\sigma$ deviation (bottom panel) of our entire $1024\times1024$ grid for different source redshifts: $z_{\rm{source}} \in \{6, 7, 10, 13\}$ in black, purple, dark blue, and light blue respectively. In the bottom panel, the $1\sigma$ deviation is plotted as a shaded region in the respective color. The strength of the absorption increases with redshift due to an increase in neutral hydrogen in the IGM. By $z_{\rm{source}}=6$, transmission peaks appear blueward of the line center and the characteristically broad red wing absorption is replaced with a sharp cutoff sensitive to the local environment around the source. Correspondingly, the increased scatter is a result of a more patchy reionization morphology during the later stages of reionization.
• Figure 3: Violin plots of the transmission $\langle e^{-\tau} \rangle$ at different redshifts for a random sample of sightlines in the thesan light-cone, assuming background quasar sources with redshifts varying between $5.529 < z < 6.542$. Each side shows two distributions, a smaller sample size of $300$ sightlines in dark blue/orange and a larger sample size of $3000$ sightlines in light blue/orange, representing current and future datasets. The left and right sides of each violin illustrate the systematic differences in employing Voigt ($\varphi$) and Dirac delta ($\delta$) absorption profiles, corresponding to $\langle e^{-\tau_\nu}\rangle$ and $\langle e^{-\tau_\delta} \rangle$, respectively. The $\delta$ model consistently results in much higher transmission, likely due to the damping-wing behavior being more accurately modeled by the Voigt Profile.
• Figure 4: The top panels show the resulting Ly$\alpha$ emission-only spectral intensity for each emission mechanism for three representative pixels in our $1024\times1024$ grid. The contributions from recombination, collisional excitation, and emission from unresolved HII regions are plotted in red, blue, and yellow, respectively, as well as the total spectral intensity in black. The middle panels show the resulting Ly$\alpha$ spectral intensity for the same three sightlines when including both emission and absorption sources. The bottom panel show the same Ly$\alpha$ spectral intensity as the middle panel, but downsampled by a factor of 100 (this sampling rate corresponds to $R\approx 52$, which is most similar to SPHEREx's $R\approx35-40$ for the relevant spectral bandpasses).
• Figure 5: The top panel shows the resulting mean of our total spectral intensity $\langle I_\nu \rangle$ for different spectral resolutions of our grid. Our nominal sampling rate $n=5000$ results in a spectral resolution $R\approx 5200$, plotted in red. Downsamplings of $10\times$ and $100\times$ are shown in violet and pale green respectively. In the bottom panel, we plot the resulting statistics of different spatial resolutions of our grid at a constant spectral resolution $R\approx 52$; the mean is plotted in violet, along with the median for different spatial downsamplings of our $1024\times1024$ grid for comparison. These downsamplings correspond to pixel scales at $z=5.5$ of $2.1"$ (the full-resolution $1024\times1024$ grid), $8.4"$, and $2.2'$, shown as red, light blue, and dark blue lines respectively. We can see that the mean is biased high but is most representative of signals extracted from Ly$\alpha$ LIM methods.