To extend the search for quasars in the epoch of reionization beyond the tip of the luminosity function, we explore point source candidates at redshift $zsim8$ in SuperBoRG, a compilation of $sim$0.4deg$^2$ archival medium-deep ($m_{rm F160W}sim 26.5$ABmag, 5$sigma$) parallel IR images taken with the Hubble Space Telescope (HST). Initial candidates are selected by using the Lyman-break technique. We then carefully analyze source morphology, and robustly identify 3 point sources at $zsim8$. Photometric redshift analysis reveals that they are preferentially fit by extra-galactic templates, and we conclude that they are unlikely to be low-$z$ interlopers, including brown dwarfs. A clear IRAC ch2 flux excess is seen in one of the point sources, which is expected if the source has strong H$beta$+[O III] emission with rest-frame equivalent width of $sim3000$AA. Deep spectroscopic data taken with Keck/MOSFIRE, however, do not reveal Ly$alpha$ emission from the object. In combination with the estimated H$beta$+[O III] equivalent width, we place an upper limit on its Ly$alpha$ escape fraction $f_{rm esc, Lyalpha}< 2 %$. We estimate the number density of these point sources $sim1times10^{-6}$Mpc$^{-3}$mag$^{-1}$ at $M_{rm UV}sim-23$mag. The final interpretation of our results remains inconclusive: extrapolation from low-$z$ studies of $faint$ quasars suggests that $>100times$ survey volume may be required to find one of this luminosity. The James Webb Space Telescope will be able to conclusively determine the nature of our luminous point source candidates, while the Roman Space Telescope will probe $sim 200$ times the area of the sky with the same observing time considered in this HST study.
Download