The AstroSat UV Deep Field South-V: Constraints on the average escape of ionizing photons in the cosmic dusk

TL;DR

The study targets LyC leakage from 49 star-forming galaxies at z ~ 1–1.5 using AstroSat UVIT F154W imaging, employing image stacking and property-based subsamples to constrain the average LyC escape. Stacks of non-detections yield no significant LyC signal, while the full sample shows only a marginal detection, indicating the LyC escape is dominated by a minority of leaker candidates. Subsample analyses reveal that low-mass, compact, high ΣSFR, blue-β galaxies contribute the strongest LyC signal, with a stacked detection around ~3σ (rising to ~4.4σ when including leakers), suggesting these conditions favor ionizing photon escape. Translating observed flux ratios into absolute escape fractions under various stellar population assumptions places ⟨fesc,abs⟩ for low-mass systems at ≈0.1–0.2 (depending on the intrinsic LyC/UV ratio) and sets stringent upper limits for the full sample (≲0.03), supporting a scenario where compact, low-mass starbursts were important but rare contributors to the ionizing background during cosmic dusk.

Abstract

We investigate the escape of ionizing (Lyman-continuum; LyC) photons from 49 star-forming galaxies at redshifts $\sim 1-1.5$, using far-ultraviolet (FUV) imaging from the Ultra-Violet Imaging Telescope (UVIT) onboard AstroSat. The sample spans a wide range of stellar masses and UV luminosities. LyC emission is undetected in most galaxies (42/49), and stacking these galaxies yields only an upper limit on the observed LyC-to-nonionizing UV flux density ratio ($({F}_{λ,\rm{LyC}}/{F}_{λ,\rm{UV}})_{\rm{obs}}<0.12$). Including all galaxies (with 7 LyC-leaker candidates) produces a marginal $2.4σ$ detection, suggesting that the average LyC signal is driven by a small number of sources. To identify the conditions favorable for LyC escape, we perform stacking analyses in bins of stellar mass, UV slope, compactness, inclination, and star formation rate surface density. A stacked LyC signal is detected at a significance of $\sim3σ$ from a subset of galaxies that are characterized as being compact, have high star formation rate surface densities, and blue UV continuum slopes, despite each of these being individually undetected in LyC. This provides the first systematic evidence at $z\sim 1-1.5$ linking these properties to LyC escape, consistent with trends observed in the lower redshift universe. Additionally, LyC leakage appears more efficient in low-mass galaxies ($\log_{10}(M_{*}/M_{\odot})<9.5$), with their average absolute escape fraction ranging from $\langle f_{\text{esc,abs}} \rangle\sim0.1-0.2$ depending on stellar population assumptions. These results support the scenario that compact, low-mass starbursts were key contributors to the ionizing photon budget during cosmic reionization.

Figures (7)

• Figure 1: Redshift distribution of our selected sample of 49 star-forming galaxies.
• Figure 2: RGB composite images of galaxies in our sample, constructed from HST imaging: F160W (red), F606W (green), and F435W (blue). Each image is displayed within a circular aperture of radius $1.4"$. The top six rows show LyC non-detections (cyan dashed apertures), while the bottom row presents the LyC leaker candidates (yellow dashed apertures) as defined in Section \ref{['sec:selection']}. The final panel shows the last LyC leaker candidate at $z\sim1.23$ (F154W ID: AUDFs_F15450), newly identified in this study, in an enlarged view, along with its observed UV continuum slope ($\beta_{\text{obs}}$), O32, and absolute UV magnitude ($M_{\text{UV}}$) values. Its UVIT F154W and N242W counterparts are highlighted within the same aperture to the right of the RGB image. Within the yellow dashed aperture (radius 1.4"), the S/N of this object is $\sim2.9$ in the F154W band image. A noise peak is evident toward the northeast in the F154W band. The F154W and N242W bands probe the LyC ($\sim690\:\text{\AA}$) and the rest-frame non-ionizing UV ($\sim1085\:\text{\AA}$) emission, respectively. The F154W band image has been smoothed with a top-hat kernel of radius 1.5 pixels to enhance visual clarity.
• Figure 3: The first two panels (from left) show the distributions of stellar mass and absolute UV magnitude for our sample: grey filled histograms represent LyC non-detections, while green hatched histograms correspond to LyC leaker candidates. In the third panel, grey (non-detections) and green (leaker candidates) star markers show the relationship between the observed UV continuum slope ($\beta_{\text{obs}}$) and absolute UV magnitude for galaxies in our sample not detected in X-rays in the Chandra 7 Ms survey Luo17. Colored circular markers indicate galaxies that are X-ray detected.
• Figure 4: Noise in the stacked images as a function of the number of images combined ($N_{\text{stack}}$). The stacked noise, $\sigma_{\text{stack}}$, is normalized by the global average noise, $\sigma$, measured in the deepest region of the F154W image Saha24, which effectively corresponds to the case of $N_{\text{stack}}=1$. Each blue point corresponds to a stacking run listed in Table \ref{['tab:stack_results']}, while the red dashed curve shows the expected $1/\sqrt{N_{\text{stack}}}$ scaling.
• Figure 5: Top panel: The F154W-band stacked image created from 42 individual $20" \times 20"$ cutouts centered on galaxies in the LyC non-detection sample. The images have been smoothed using a Gaussian kernel having a standard deviation of nearly 1 pixel. $N_{\text{stack}}$ denotes the number of galaxies contributing to the stacked image. Right panel: Normalized distribution of the normalized fluxes measured within randomly placed apertures across the stacked image, excluding the central region. The fluxes are normalized by the standard deviation of a Gaussian fitted to the distribution of fluxes, denoted here by the solid red curve. The dashed blue line indicates the significance of the flux measured within the central white dashed aperture. Bottom panel: Same as above, but for the stack of F154W cutouts corresponding to the 7 LyC leaker candidates.