CLASSY XIII. Cutting through the Clouds - Comparing Indirect Tracers of Ionizing Photon Escape

TL;DR

This study critically evaluates six empirically calibrated indirect tracers of ionizing photon escape using the CLASSY galaxy sample, combining LIS covering fractions, UV slopes, Lyα kinematics, a multivariate LyCsurv model, radiation-hydrodynamic mock spectra, and the O32 ratio. By comparing these diagnostics and their medians, the paper reveals substantial method-to-method variance but broadly consistent leaker classifications, with about half the galaxies predicted to have $\left< f_{ m esc}^{\rm LyC} \right> > 1\%$ and evidence for two LyC-escape pathways: an early escape tied to very young stellar populations and a delayed escape linked to SN-driven ISM clearing. The analysis highlights the complementary strengths and biases of different tracers, showing strong ties between dust and LyC attenuation in some cases and line-of-sight neutral-gas effects in others, and emphasizes a multi-tracer approach for robustly inferring LyC escape relevant to reionization. These results imply that understanding the diversity of LyC-escape mechanisms is essential for accurately modeling the ionizing emissivity history of the universe during the Epoch of Reionization.

Abstract

The Epoch of Reionization (EoR) provides critical insights into the role of early galaxies in shaping the ionization state of the universe. However, because of the opacity of the intergalactic medium, it is often not possible to make direct measurements of the ionizing photon escape fraction ($f_{\mathrm{esc}}^{\: \mathrm{LyC}}$) of high-redshift ($z \gtrsim 4$) galaxies. To explore the agreement and systematics of common indirect approaches, we applied six empirically calibrated diagnostics to predict $f_{\mathrm{esc}}^{\: \mathrm{LyC}}$ for the 45 nearby star-forming galaxies from the COS Legacy Spectroscopic SurveY (CLASSY). These methods- based on ultraviolet (UV) absorption lines, the UV continuum slope, Ly$α$ kinematics, a multivariate model, radiation-hydrodynamic simulations, and nebular emission line ratios- enable us to explore systematic differences between predictions and assess how galactic properties influence inferred LyC escape. Despite significant variations in method predictions, there is broad consistency in the resulting weak and strong LyC leaker classifications, with approximately half exhibiting predicted escape fractions $>$1%. We find evidence for two different pathways of LyC escape in nearby star-forming galaxies: (1) an early escape model driven by very young stellar populations, and (2) a delayed escape model that is consistent with supernova-driven outflows and time-dependent ISM clearing. The early escape model is favored among galaxies with a single, intense burst of recent star formation. In contrast, the delayed escape model is common among galaxies with more extended starburst histories. To interpret ionizing photon escape during the EoR, it will be necessary to recognize and understand this diversity in LyC escape mechanisms.

Figures (10)

• Figure 1: Example fits for one CLASSY galaxy, J1129+2034. Top: the Starburst99Leitherer_1999 stellar population fit and H$\;$ fit (orange), which results in a light-weighted metallicity of Z/Z$_\odot$ = 0.21 and light-weighted age of 2.56 Myr for this galaxy. Bottom: the $_{\mathrm{UV}}$-slope fit (dashed black line with cyan MC uncertainties). In both panels, the observed spectrum is in black with purple uncertainties. The windows used in each fit are shaded in light blue and light orange in the top and bottom panels, respectively. The windows for the SPS fits were chosen to exclude regions with strong nebular features (such as the ISM absorption component of C$\;$$$$$1549,1550). In addition, in galaxies with broad He$\;$$$1640 emission, we masked out this feature since SB99 has does not include sufficient Wolf-Rayet star templates to reproduce this broad emission, especially at low-metallicities and/or young ages Leitherer_2018Chisholm_2019.
• Figure 2: The trend between direct measurements of $f_{\mathrm{esc}}^{\: \mathrm{LyC}}$ and the peak separation of Ly$$, $\mathrm{v}_{\mathrm{sep}}^{\mathrm{Ly} }$. Individual measurements of $\mathrm{v}_{\mathrm{sep}}^{\mathrm{Ly} }$ from Leitet_2013, Borthakur_2014, and Leitherer_2016, with escape fractions derived by Chisholm_2017, are shown by empty circles. Measurements from Izotov_2016a and Izotov_2016b are shown with empty stars, while those from Izotov_2018a and Izotov_2018b are filled stars. Lastly, the measurements from Flury_2022a for the eight LzLCS+ galaxies with double-peaked Ly$$ emission are shown by filled circles. The best-fit curve for these points is listed in Equation \ref{['eq:vsep']} and is in close agreement with the one reported by Izotov_2018b.
• Figure 3: $f_{\mathrm{esc}}^{\: \mathrm{LyC}}$ from different indirect methods for CLASSY galaxies, as shown by different shapes of empty markers: diamonds ($f_{\mathrm{esc}}^{\: C_f}$), vertical triangles ($f_{\mathrm{esc}}^{\: }$), pluses ($f_{\mathrm{esc}}^{\: v_{\mathrm{sep}}}$), squares ($f_{\mathrm{esc}}^{\: \mathrm{O}_{32}}$), horizontal triangles ($f_{\mathrm{esc}}^{\mathrm{sim}}$), and hexagons ($f_{\mathrm{esc}}^{\: \mathrm{AFT}}$). The medians of these estimates are shown by purple points, with uncertainties represented by the 16th and 84th percentiles from 300 MC variations of the individual $f_{\mathrm{esc}}^{\: \mathrm{LyC}}$ predictions based on their uncertainties. Overall, $\sim50\%$ of CLASSY galaxies have $\left<f_{\mathrm{esc}}^{\: \mathrm{LyC}}\right> > 1\%$, with multiple meeting our criterion for weak LyC leakers ($5\% \leq \left<f_{\mathrm{esc}}^{\: \mathrm{LyC}}\right> < 20\%$) and one that is a candidate for strong LyC leaking (J1323-0132; $\left<f_{\mathrm{esc}}^{\: \mathrm{LyC}}\right> \geq 20\%$).
• Figure 4: Pairwise comparison of six indirect estimates of $f_{\mathrm{esc}}^{: \mathrm{LyC}}$ for the CLASSY sample, as well as the median of these predictions $\left<f_{\mathrm{esc}}^{\: \mathrm{LyC}}\right>$. Each panel displays the Kendall’s tau ($_K$) correlation coefficient between two methods, with the points color-coded by the $_K$ value to visually highlight the strength of the correlation. Strongest correlations are found between methods that depend on shared observables (e.g., LIS absorption and dust), while weaker correlations appear for physically distinct methods such as O$_{32}$ or Ly$$ kinematics. This figure emphasizes both the areas of internal agreement and the systematic differences among the predictors.
• Figure 5: The difference between each indirect $f_{\mathrm{esc}}^{\: \mathrm{LyC}}$ measurement and the average of the other five indirect methods $\left<f_{\mathrm{esc}}^{\: \mathrm{LyC}}\right>^*$ as a function of $E(B-V)_)$ for the left column and the C$\;$$$1334 residual flux for the right column. The $y$-axes correspond to how closely each estimate agrees with the other five, considering their uncertainties $$. This figure shows how different methods may be biased based on the amount of dust (left) and/or the amount of neutral gas (right) are present along the line-of-sight. Each panel includes a linear fit (dashed line) and lists the Kendall's tau correlation coefficient and corresponding $p$-value. The twelve galaxies with $C_f$(H$\;$) $< 0.9$ are shown as upper limits in the first row.