f_esc

Code accompanying the paper:

Ionizing photon production and escape fractions during cosmic reionization in the TNG50 simulation Ivan Kostyuk, Dylan Nelson, Benedetta Ciardi, Martin Glatzle, Annalisa Pillepich MNRAS 521, 3077 (2023) — DOI: 10.1093/mnras/stad677 · arXiv:2207.11278

Science context

Understanding which galaxies supply the bulk of ionizing photons during the epoch of reionization (6 < z < 10) requires knowing not just how many ionizing photons are produced, but what fraction escape from their host halos. Measuring this escape fraction f_esc directly at high redshift is observationally out of reach, and theoretical predictions differ substantially depending on how radiative transfer and sub-grid absorption are modelled.

This repository contains the analysis code for a study that post-processes ~4 000 TNG50 galaxies (10⁶ ≤ M_★/M_☉ ≤ 10⁸) with the 3D multi-frequency radiative transfer code CRASH, computing f_esc self-consistently along thousands of lines of sight per halo.

Key results:

• f_esc rises with stellar mass from ~0.3 at M_★ = 10⁶ M_☉ to ~0.6 at M_★ = 10^7.5 M_☉, with hints of a turnover at higher masses
• Significant scatter at fixed mass, driven by diversity in the ionizing photon rate and the spatial relationship between stellar sources and the gas density field
• Dust reduces f_esc by a few percent at low masses and up to 10% for M_★ ≳ 10^6.5 M_☉
• f_esc is energy-dependent: photons above 54.4 eV (relevant for He reionization and binary-star models) escape at a substantially lower rate than lower-energy ionizing photons
• Halos with M_★ ≲ 10^7.5 M_☉ dominate the global ionizing emissivity at all redshifts studied

Repository structure

The notebooks in this repo are load-bearing: each one produces specific figures in the paper and contains the full analysis pipeline for that result. They are not exploratory scratch pads — they are the analysis.

Python scripts

These scripts interact directly with the CRASH post-processing runs or TNG50 snapshot data and must be run before the analysis notebooks.

Script	Purpose
build_df.py	Reads CRASH output folders and assembles a pandas dataframe of halo properties
update_df.py	Adds stellar ages, gas clumping, and surface density columns to the dataframe
add_metallicities.py	Appends stellar metallicities to the dataframe
prepare_new_runs.py	Tools for setting up a new CRASH post-processing run
synchronize_folders.py	Moves halos from reduced-source RT runs to full-source fesc calculations
phase_diagram.py	Generates phase diagram data

Analysis notebooks (paper figures)

These notebooks read the compiled dataframe and produce the paper figures. They can be run independently of the simulation data, provided the dataframe is available (see Data below).

Notebook	Figures	What it does
counts_histogram.ipynb	Fig. 1	Halo population summary plots
median_fesc.ipynb	Figs. 2, 3, 13	fesc vs stellar/halo mass; dust comparison; literature comparison
fesc_vs_quant.ipynb	—	fesc as a function of various galaxy properties
histograms.ipynb	Fig. 6	2D histograms of halo properties vs fesc; bimodal escape fraction analysis
loc_esc.ipynb	Fig. 7	Effect of different local (cloud-scale) escape fraction assumptions
uv_emissivity.ipynb	Figs. 8, 9	Escaped ionizing photon density in TNG50; comparison to literature
r_fesc.ipynb	Fig. 10	Properties of individual stellar particles and their contribution to fesc
spectra.ipynb	Figs. 11, 12	Spectral (energy) dependence of fesc
numerical_convergence_tests.ipynb	Fig. A1	Source number reduction convergence tests
large_radii.ipynb	—	fesc as a function of aperture radius (not in final paper)

Notebooks requiring full simulation data

These two notebooks need the raw CRASH density and ionization maps, not just the dataframe:

Notebook	Figures	What it does
esc_fraction.ipynb	Fig. 5	Line-of-sight fesc maps and density projections for individual halos
halo_image.ipynb	Fig. 4	Projected images of halo gas and stellar distributions

Simulation management notebooks

Not used for data analysis — for preparing and cleaning CRASH runs:

Notebook	Purpose
merge_sources.ipynb	Source reduction pre-processing; includes validation tests
clean_up.ipynb	Deletes intermediate simulation files to reduce disk usage

Data

The analysis notebooks require the compiled halo dataframe, which is not included in this repository due to size. To obtain the dataframe, please open an issue or reach out via the contact on the arXiv page.

Requirements

Python ≥ 3.7, NumPy, pandas, matplotlib, h5py, astropy. Access to IllustrisTNG simulation data and CRASH outputs is required to run the simulation-facing scripts.

Citation

If you use this code, please cite:

@article{Kostyuk2023fesc,
  author  = {Kostyuk, Ivan and Nelson, Dylan and Ciardi, Benedetta and Glatzle, Martin and Pillepich, Annalisa},
  title   = {Ionizing photon production and escape fractions during cosmic reionization in the {TNG50} simulation},
  journal = {MNRAS},
  year    = {2023},
  volume  = {521},
  pages   = {3077},
  doi     = {10.1093/mnras/stad677},
  eprint  = {2207.11278},
  archivePrefix = {arXiv}
}