Disentangling the galactic and intergalactic components in 313 observed Lyman-alpha line profiles between redshift 0 and 5

TL;DR

This work presents zELDA, a neural-network–based framework to disentangle galactic (ISM/CGM) and intergalactic (IGM) contributions to Lyα line profiles. By training three ANN models on mock spectra and using IGM transmissions from the TNG100 simulation, the authors reconstruct intrinsic galactic Lyα spectra and quantify the Lyα IGM escape fraction across $0\lesssim z\lesssim5$ for 313 observed lines from HST/COS and MUSE. They show that the intrinsic galactic Lyα profile undergoes little evolution with redshift once IGM effects are removed, while the IGM attenuation, particularly of the blue wing, grows with redshift and drives the observed spectral evolution. The results yield a redshift–dependent mean IGM escape fraction $\langle f_{ m esc}^{4\AA} \rangle$ that declines from ~0.9 at $z\sim3$ to ~0.6 by $z\sim5$, consistent with mean IGM transmission studies, and they demonstrate the potential for per-source deconvolution and three-dimensional IGM tomography using Lyα spectra.

Abstract

Lyman-Alpha (Lya) photons emitted in star-forming regions inside galaxies experience a complex radiative transfer process until they reach the observer. The Lya line profile that we measured on Earth is, thus, the convolution of the gas properties in the interstellar (ISM), circumgalactic (CGM), and intergalactic medium (IGM). We make use of the open source package zELDA (redshift Estimator for Line profiles of Distant Lyman-Alpha emitters) to disentangle the galactic and IGM components of the Lya profiles to study both the evolution of the intrinsic galactic emission and the IGM transmission across cosmic time. zELDA includes different artificial neural networks that reconstruct IGM attenuated Lya line profiles. These models are trained using mock Lya line profiles. A Monte Carlo radiative transfer code computes the galactic component for the so-called thin shell model. Moreover, the IGM component is included through the IGM transmission curves generated from the IllustrisTNG100 cosmological galaxy formation simulation. We recover their intrinsic galactic spectra by applying the zELDA to 313 Lya line profiles observed with HST/COS and MUSE. Sources at z < 0.5 show weak IGM attenuation, while at z > 3, ZELDA reveals significant IGM suppression of the blue peak in several sources. After separating the IGM effects, the stacked intrinsic galactic Lya profiles show a minimal evolution from z = 0 to 6. The mean IGM transmission for z < 0.5 in HST/COS data exceeds 90%, while the MUSE data show an evolution from 0.85 at z = 3.0 to 0.55 at z = 5.0.

Figures (26)

• Figure 1: Illustration of the impact of the IGM attenuation in the galactic ${\rm Ly}\alpha$ line profile. Through all the panels, we show the intrinsic ${\rm Ly}\alpha$ line profile emerging from the galaxy in thin black and the IGM transmission curve in yellow. In the top panels we show the galactic line profile obscured intrinsically (top left, purple), the same profile after traveling a short distance in the IGM (blue, center left) as well as further through the IGM (green, top right), and its shape when it reaches the observer (red). The bottom panel show the zELDA reconstruction (purple) using as input the observed line profile (red), along with the line structure terminology used in this work.
• Figure 2: Nine examples of zELDA's prediction on line profiles displayed in the rest frame. The observed line profile is shown in dark grey with its $1\sigma$ uncertainty in light grey. zELDA's reconstruction using the models IGM+z, IGM-z, and NoIGM is displayed in blue, green, and yellow, respectively. The redshift of the source is shown in the top right corner. zELDA's estimations of $f_{\rm esc}^{4\AA}$ given by the IGM+z and IGM-z models are shown in blue and green, respectively. The central grey-shaded region shows the wavelength interval of $f_{\rm esc}^{4\AA}$. The estimated outflow parameters are indicated in the table with each subplot. $V_{\rm exp}$ is given in km s${}^{-1}$, $N_{\rm H}$ in $cm^{-2}$ and $EW_{\rm in}$ and $W_{\rm in}$ in Å. All the fitted line profiles are also shown in Appendix \ref{['app:lines']}.
• Figure 3: Spearman correlation coefficient and its 1$\sigma$ associated uncertainty for HST and MUSE sources ,for IGM+z (bottom blue) and IGM-z (top green).
• Figure 4: $f_{\rm esc}^{4\AA}$ as a function of redshift for IGM+z (left) and IGM-z (middle). The right panel shows the histogram of the predicted $f_{\rm esc}^{4\AA}$ for IGM+z and IGM-z in blue an green, respectively.
• Figure 5: Mean $f_{\rm esc}^{4\AA}$ as a function of redshift (squares) for IGM+z (blue) and IGM-z (green). The solid lines indicate the fits, including all the sources below $z=5$. Meanwhile, the colored dashed line indicates the fit using only MUSE data below $z=5$.