pith. sign in

arxiv: 2503.03824 · v4 · submitted 2025-03-05 · 🌌 astro-ph.GA

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

Pith reviewed 2026-05-23 00:52 UTC · model grok-4.3

classification 🌌 astro-ph.GA
keywords Lyman-alphaintergalactic mediumescape fractiongalactic outflowsredshift evolutionradiative transferneural networksline profiles
0
0 comments X

The pith

After removing intergalactic effects, the intrinsic Lyman-alpha line profiles emitted by galaxies show little change from redshift 0 to 6.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper uses a neural-network tool to separate the contributions of gas inside galaxies from absorption by the intergalactic medium in 313 observed Lyman-alpha spectra spanning redshifts 0 to 6. Once the intergalactic contribution is subtracted, the stacked profiles produced inside galaxies remain nearly constant across this entire range. The fraction of Lyman-alpha photons that pass through the intergalactic medium falls from above 90 percent at the lowest redshifts to roughly 55 percent by redshift 5. The measurements line up with independent estimates of intergalactic opacity and indicate that the intergalactic medium becomes the dominant obstacle to observing Lyman-alpha above redshift 5, while processes inside galaxies matter more at lower redshifts.

Core claim

Using artificial neural networks trained on mock Lyman-alpha spectra generated with Monte Carlo radiative transfer through thin-shell models and intergalactic transmission curves, the method disentangles galactic and intergalactic contributions in 313 observed line profiles. After correcting for intergalactic effects, the stacked intrinsic galactic Lyman-alpha line profiles display remarkably little evolution from z=0 to z=6. The mean intergalactic Lyman-alpha escape fraction exceeds 90 percent for z<0.5, decreases from approximately 0.85 at z=3 to approximately 0.55 at z=5, and agrees with independent constraints on the intergalactic mean optical depth; this implies that the intergalactic介质

What carries the argument

The neural-network tool trained on mock spectra from thin-shell galactic models combined with intergalactic transmission curves, which separates the galactic emission component from intergalactic absorption in observed Lyman-alpha profiles.

If this is right

  • The mean intergalactic escape fraction decreases steadily with increasing redshift.
  • Intergalactic absorption suppresses the blue peak of the line profile at redshifts above 3.
  • At redshifts greater than or equal to 5 the intergalactic medium dominates the limitation on Lyman-alpha observability.
  • At lower redshifts the interstellar and circumgalactic media become the primary factors controlling the observed line strength.
  • The reconstructed intrinsic profiles can be compared directly with global escape-fraction measurements to isolate the separate roles of galactic and intergalactic gas.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The reported lack of evolution in intrinsic profiles would constrain models of how galactic outflows change with cosmic time if the separation remains accurate at higher redshifts.
  • Applying the same separation technique to larger samples from upcoming surveys could map spatial variations in intergalactic transmission on megaparsec scales.
  • If the neural-network reconstruction proves robust, it offers a route to calibrate the Lyman-alpha transmission in reionization simulations against direct observational constraints.
  • The method could be tested on local galaxies where intergalactic effects are negligible to quantify any systematic bias introduced by the training assumptions.

Load-bearing premise

The neural networks trained on thin-shell models and the chosen intergalactic transmission curves can accurately separate galactic and intergalactic effects when applied to real observed spectra.

What would settle it

Independent measurements of the intergalactic Lyman-alpha escape fraction at z approximately 5 that differ substantially from 0.55 using a method that does not rely on the same thin-shell models would falsify the reported redshift evolution.

read the original abstract

Lyman-Alpha (Lya) photons emitted in star-forming galaxies undergo complex radiative transfer through the interstellar (ISM), circumgalactic (CGM), and intergalactic medium (IGM), imprinting characteristic signatures on their observed line profiles. We use the open-source package zELDA (redshift Estimator for Line profiles of Distant Lyman-Alpha emitters) to disentangle the galactic and intergalactic contributions in 313 Lya spectra observed with HST/COS and MUSE, spanning 0<z<6. zELDA employs artificial neural networks trained on mock Lya spectra generated with Monte Carlo radiative transfer through thin-shell models and IGM transmission curves from the TNG100 simulation. We find that sources at $z<0.5$ exhibit minimal IGM attenuation, whereas at $z>3$ the IGM significantly suppresses the blue peak of Lya. After correcting for IGM effects, the stacked intrinsic galactic Lya line profiles display remarkably little evolution from $z=0$ to $z=6$. We measure the mean IGM Lya escape fraction, finding $<f^{IGM}_{esc}> > 90\%$ for z<0.5, decreasing from $\sim0.85$ at $z=3$ to $\sim0.55$ at $z=5$. Our measurement of the redshift evolution of the Lya IGM escape fraction agrees with independent constraints on the IGM mean optical depth. After a comparison between our $<f^{IGM}_{esc}>$ estimation and the global Lya escape fraction from the literature, our findings indicate that the IGM might dominate Lya observability at redshift z$\gtrsim$5.0, after which ISM and CGM effects tend to dominate at lower $z$. Our results demonstrate that zELDA enables robust reconstruction of intrinsic Lya spectra and provides a direct probe of the interplay between galactic outflows and IGM transmission across cosmic time.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript applies the zELDA neural network (trained on thin-shell Monte Carlo radiative transfer mocks plus TNG100 IGM transmission curves) to disentangle galactic and intergalactic contributions in 313 HST/COS and MUSE Lyman-alpha spectra spanning 0<z<6. After IGM correction the stacked intrinsic galactic profiles are reported to show remarkably little redshift evolution; the mean IGM escape fraction is measured as >90% below z=0.5, falling from ~0.85 at z=3 to ~0.55 at z=5, and stated to agree with independent IGM optical-depth constraints. The work concludes that IGM effects dominate Lya observability at z≳5 while galactic effects dominate at lower redshift.

Significance. If the disentangling is robust, the result supplies a direct empirical separation of galactic versus IGM contributions across cosmic time and quantifies the redshift evolution of f_esc^IGM, with potential implications for interpreting high-z Lya surveys. The open-source zELDA package and the use of a sizable observational sample are positive features.

major comments (2)
  1. [Methods (ANN training)] Methods section (ANN training description): the networks are trained exclusively on thin-shell Monte Carlo RT models combined with TNG100 IGM curves. Thin-shell geometries impose a specific velocity-density structure that may not span the diversity of real ISM/CGM configurations (clumpy media, multi-phase outflows, extended halos). This assumption is load-bearing for the central claims, because any systematic mismatch can cause the network to attribute blue-peak suppression or asymmetry to IGM attenuation rather than galactic RT, thereby biasing the recovered intrinsic stacks and the reported f_esc^IGM(z) values toward the claimed flat evolution.
  2. [Results (f_esc^IGM comparison)] Results (IGM escape fraction and comparison paragraph): the claim that <f_esc^IGM> agrees with independent IGM optical-depth constraints requires an explicit demonstration that the comparison is not circular with the TNG100 training data; without this, the robustness of the redshift trend cannot be fully assessed.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'remarkably little evolution' should be replaced by a quantitative statement (e.g., change in peak velocity or FWHM between redshift bins) with a figure reference.
  2. [Figures] Figure captions and axis labels: ensure every panel explicitly distinguishes observed, IGM-corrected intrinsic, and model IGM transmission profiles.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify key aspects of our methodology and results. We address each major comment below and will revise the manuscript accordingly to improve transparency on model assumptions and the independence of our comparisons.

read point-by-point responses
  1. Referee: Methods section (ANN training description): the networks are trained exclusively on thin-shell Monte Carlo RT models combined with TNG100 IGM curves. Thin-shell geometries impose a specific velocity-density structure that may not span the diversity of real ISM/CGM configurations (clumpy media, multi-phase outflows, extended halos). This assumption is load-bearing for the central claims, because any systematic mismatch can cause the network to attribute blue-peak suppression or asymmetry to IGM attenuation rather than galactic RT, thereby biasing the recovered intrinsic stacks and the reported f_esc^IGM(z) values toward the claimed flat evolution.

    Authors: We agree that thin-shell models represent a simplification and do not encompass the full range of possible ISM/CGM structures. These models have, however, been shown in the literature to reproduce the dominant observed Lyα profile features (double-peaked shapes, peak ratios) across a wide parameter space of outflow velocity, HI column density, and dust content. Our training grid spans this space densely. To address the concern directly, we will add an explicit discussion subsection in Methods on the limitations of the thin-shell geometry, including potential biases in attributing blue-peak suppression, and note that future extensions could incorporate clumpy or multi-phase mocks. This revision will not alter the reported results but will better contextualize them. revision: yes

  2. Referee: Results (IGM escape fraction and comparison paragraph): the claim that <f_esc^IGM> agrees with independent IGM optical-depth constraints requires an explicit demonstration that the comparison is not circular with the TNG100 training data; without this, the robustness of the redshift trend cannot be fully assessed.

    Authors: The training employs TNG100-derived transmission curves, but the comparison paragraph references independent observational constraints on the IGM effective optical depth τ_eff (e.g., from high-resolution quasar spectra in the literature, such as those compiled in studies of the Lyα forest at z~3–6). These are not outputs of TNG100. We will revise the text to explicitly name the observational sources, state that the agreement is with these external data, and add a short sentence clarifying the distinction between the simulation-based training set and the observational validation. This removes any ambiguity about circularity. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation applies trained model to external data

full rationale

The paper trains ANNs on thin-shell Monte Carlo RT mocks plus TNG100 IGM transmission curves, then applies the resulting zELDA network to 313 observed HST/COS and MUSE spectra. The reported intrinsic stacked profiles and <f^IGM_esc>(z) are outputs of this forward application, not inputs. The abstract explicitly states agreement with independent IGM optical depth constraints, providing an external benchmark. No quoted step shows a fitted parameter renamed as prediction, a self-definitional loop, or a load-bearing self-citation that reduces the central claim to its own assumptions. The derivation chain remains self-contained against the cited external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; main unverified premise is fidelity of thin-shell + TNG100 training data to real radiative transfer. No free parameters or invented entities are identifiable from the abstract.

axioms (1)
  • domain assumption Thin-shell models and TNG100 IGM curves produce mock spectra sufficiently representative for NN training to generalize to observations
    Explicitly stated as the basis for generating training data in the abstract.

pith-pipeline@v0.9.0 · 5949 in / 1259 out tokens · 33215 ms · 2026-05-23T00:52:11.154475+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.