Ionizing Photon Escape Fraction

• Ionizing Photon Escape Fraction is the proportion of hydrogen-ionizing photons escaping from galaxies, defined by the ratio of emergent to intrinsic stellar outputs.
• It depends on halo mass, gas distribution, and star formation geometry, with simulations showing values from <1% in massive halos to 10–40% in low-mass systems.
• Temporal variability and anisotropic escape, driven by feedback and ISM clumping, play a crucial role in cosmic reionization modeling.

The ionizing photon escape fraction, typically denoted $f_{\rm esc}$, quantifies the proportion of hydrogen-ionizing (Lyman continuum; $\lambda < 912$ Å) photons produced by stellar populations in galaxies that escape into the intergalactic medium (IGM), thereby contributing to the reionization of cosmic hydrogen. The determination and modeling of $f_{\rm esc}$ is pivotal for understanding whether observed galaxies provide sufficient ionizing photons to explain the timing and morphology of reionization, as constrained by observations such as the Thomson optical depth, Lyman alpha emitters, and high-redshift quasar damping wings.

1. Formal Definition and Physical Basis

$f_{\rm esc}$ is formally defined for a galaxy or halo as: $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$

where $N_{\rm emitted}$ denotes the total number of produced ionizing photons and $N_{\rm phot}(r \geq r_{200})$ counts those emerging beyond the virial radius (Paardekooper et al., 2015). This can equivalently be expressed in terms of luminosity,

$f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$

with $L_{\rm esc}$ the emergent ionizing luminosity and $L_{\rm int}$ the intrinsic value.

At the population level: $\lambda < 912$0 where $\lambda < 912$1 is the UV luminosity density, $\lambda < 912$2 is the production efficiency, and $\lambda < 912$3 encodes halo mass and redshift dependence (Finkelstein et al., 2019).

2. Physical Drivers and Dependencies

Halo Mass Dependence

Radiation hydrodynamics simulations and radiative transfer post-processing demonstrate a strong, nonlinear halo mass dependence (Paardekooper et al., 2015, Yajima et al., 2010). For halos with virial mass below $\lambda < 912$4 M$\lambda < 912$5, shallow potential wells allow SNe and massive-star feedback to clear low column-density channels near young clusters, yielding $\lambda < 912$6–40%. By contrast, in $\lambda < 912$7–$\lambda < 912$8 M$\lambda < 912$9 systems, dense gas and deep central embedding of star-forming regions reduce $f_{\rm esc}$0 to below 1%.

Summarized scaling from simulations:

$f_{\rm esc}$1 [$f_{\rm esc}$2]	$f_{\rm esc}$3
$f_{\rm esc}$4	$f_{\rm esc}$5
$f_{\rm esc}$6	$f_{\rm esc}$7
$f_{\rm esc}$8	$f_{\rm esc}$9

In the “First Billion Years” project, the escape fraction probability density function $f_{\rm esc}$0 is broad in each mass bin, with up to $f_{\rm esc}$11 dex scatter (Paardekooper et al., 2015, Yajima et al., 2010).

Gas Distribution and Star Formation Geometry

The local column density of neutral gas within $f_{\rm esc}$210 pc of star clusters, $f_{\rm esc}$3, is the principal constraint in high-$f_{\rm esc}$4 halos (Paardekooper et al., 2015). Porosity of the ISM, number and density of clumps, and the offset of young clusters from the neutral gas centroid modulate $f_{\rm esc}$5. Analytical and simulation models (Fernandez et al., 2010) show that fewer, denser clumps yield higher escape fractions due to the decreased probability of intercepting a sightline through each clump.

3. Temporal and Angular Anisotropy

Due to bursty and clustered star formation, and delayed feedback effects, $f_{\rm esc}$6 is highly time-variable on $f_{\rm esc}$7Myr timescales (Kimm et al., 2014). SNe typically create low-density escape channels some $f_{\rm esc}$810 Myr after the star formation peak, producing brief intervals when the instantaneous escape fraction can exceed 20%, though the photon-weighted average over time is lower (typically $f_{\rm esc}$911–14% depending on feedback and IMF) (Kimm et al., 2014).

The angular escape fraction distribution is strongly anisotropic with most photons escaping through narrow “beams” (few steradian cones) (Paardekooper et al., 2015). In halos with $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$0, the escape solid angle is $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$1 sr, but if $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$2, typical escape cones cover $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$3 sr.

4. Population-Averaged, Luminosity, and Redshift Dependence

To reconcile low mean escape fractions with global reionization constraints, Finkelstein et al. (Finkelstein et al., 2019) propose a halo-mass–dependent $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$4, scaled by a global factor ($$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$5, 68% CI 3.3–7.5), allowing galaxies down to the atomic cooling limit to contribute. Posterior-averaged population escape fractions span:

• $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$6 at $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$7
• $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$8 at $$f_{\rm esc} = \frac{N_{\rm phot}(r \geq r_{200})}{N_{\rm emitted}}$$9
• $N_{\rm emitted}$0 by $N_{\rm emitted}$1

Fainter galaxies ($N_{\rm emitted}$2) reach $N_{\rm emitted}$3–12% at $N_{\rm emitted}$4, while brighter ($N_{\rm emitted}$5) systems yield only $N_{\rm emitted}$61–3%. These trends are crucial, as the steep faint-end slope of galaxy LFs at high $N_{\rm emitted}$7 ensures that faint objects dominate the ionizing photon budget (Finkelstein et al., 2019).

5. Observational Constraints and Tension

Population-averaged escape fractions are matched to key observables:

• Becker & Bolton (2013) emissivity at $N_{\rm emitted}$8–4.75
• Planck optical depth $N_{\rm emitted}$9, with models achieving $N_{\rm phot}(r \geq r_{200})$0 (a $N_{\rm phot}(r \geq r_{200})$1 offset)
• McGreer et al. (2015) $N_{\rm phot}(r \geq r_{200})$2: reionization completes by $N_{\rm phot}(r \geq r_{200})$3, midpoint $N_{\rm phot}(r \geq r_{200})$4, $N_{\rm phot}(r \geq r_{200})$5 at $N_{\rm phot}(r \geq r_{200})$6, mildly ($N_{\rm phot}(r \geq r_{200})$71–2$N_{\rm phot}(r \geq r_{200})$8) in tension with Ly$N_{\rm phot}(r \geq r_{200})$9 emitter and quasar damping wing constraints which favor $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$0–0.6 (Finkelstein et al., 2019).

Models with a single $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$1 at all redshifts/luminosities generally underproduce the observed ionizing emissivity, reinforcing the necessity of mass- and redshift-dependent escape fractions.

6. Role of AGN and Other Ionizing Sources

In Finkelstein et al. (Finkelstein et al., 2019), AGN contribute a non-negligible ($f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$230% at $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$3) but subdominant component to the overall ionizing budget, never dominating before $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$4. Parameter posterior constraints yield an AGN scale factor of $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$5 and a redshift slope of $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$6 ($f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$7 to $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$8) for the AGN term.

7. Implications for Reionization Modeling and Physical Interpretation

The dominance of ultra-faint dwarfs in the ionizing budget, broad stochasticity in $f_{\rm esc} = \frac{L_{\rm esc}}{L_{\rm int}}$9, and strong anisotropy and time variability present significant modeling challenges. Semi-analytic and cosmological-volume reionization codes are advised to adopt mass-dependent and probabilistic $L_{\rm esc}$0 distributions rather than deterministic or uniform prescriptions (Paardekooper et al., 2015). In practical terms, power-law or broken-power-law $L_{\rm esc}$1 relations with log-normal scatter are supported by simulation results (see schematic scaling in Section 2).

The requirement for low mean escape fractions ($L_{\rm esc}$2) to suffice for reionization is satisfied only if galaxies form stars to the atomic cooling limit pre-reionization and photosuppression mass $L_{\rm esc}$3 is invoked after (Finkelstein et al., 2019).

Finally, physical processes—including supernova feedback, ISM clumpiness, geometry of star cluster embedding, and rising ionizing photon production efficiency at higher $L_{\rm esc}$4/fainter $L_{\rm esc}$5—are crucial to a realistic understanding of $L_{\rm esc}$6.

---

In summary, contemporary models and simulations indicate that a low average escape fraction ($L_{\rm esc}$7) can reionize the universe, but only via a sharply mass-dependent $L_{\rm esc}$8 scaling favoring low-mass, ultra-faint galaxies, with significant contributions from small-scale ISM structure, strong anisotropy, temporal stochasticity, and secondary AGN emission, matching most observational constraints to within $L_{\rm esc}$9 (Finkelstein et al., 2019, Paardekooper et al., 2015, Yajima et al., 2010).