A High-Resolution Hubble Space Telescope Study of Apparent Lyman Continuum Leakers at $z\sim3$

We present $U_{336}V_{606}J_{125}H_{160}$ follow-up $HST$ observations of 16 $z\sim3$ candidate LyC emitters in the HS1549+1919 field. With these data, we obtain high spatial-resolution photometric redshifts of all sub-arcsecond components of the LyC candidates in order to eliminate foreground contamination and identify robust candidates for leaking LyC emission. Of the 16 candidates, we find one object with a robust LyC detection that is not due to foreground contamination. This object (MD5) resolves into two components; we refer to the LyC-emitting component as MD5b. MD5b has an observed 1500Å to 900Å flux-density ratio of $(F_{UV}/F_{LyC})_{obs}=4.0\pm2.0$, compatible with predictions from stellar population synthesis models. Assuming minimal IGM absorption, this ratio corresponds to a relative (absolute) escape fraction of $f_{esc,rel}^{MD5b}=75-100$% ($f_{esc,abs}^{MD5b}=14-19$%). The stellar population fit to MD5b indicates an age of $\lesssim50$Myr, which is in the youngest 10% of the $HST$ sample and the youngest third of typical $z\sim3$ Lyman break galaxies, and may be a contributing factor to its LyC detection. We obtain a revised, contamination-free estimate for the comoving specific ionizing emissivity at $z=2.85$, indicating (with large uncertainties) that star-forming galaxies provide roughly the same contribution as QSOs to the ionizing background at this redshift. Our results show that foreground contamination prevents ground-based LyC studies from obtaining a full understanding of LyC emission from $z\sim3$ star-forming galaxies. Future progress in direct LyC searches is contingent upon the elimination of foreground contaminants through high spatial-resolution observations, and upon acquisition of sufficiently deep LyC imaging to probe ionizing radiation in high-redshift galaxies.