Hintze Centre for Astrophysical Surveys

Abstract:

Sub-milllimetre galaxies (SMGs) are some of the most luminous star-forming galaxies in the Universe, however their properties remain hard to determine due to the difficulty of identifying their optical\slash near-infrared counterparts. One of the key steps to determining the nature of SMGs is measuring a redshift distribution representative of the whole population. We do this by applying statistical techniques to a sample of 761 850$\mu$m sources from the SCUBA-2 Cosmology Legacy Survey observations of the UKIDSS Ultra-Deep Survey (UDS) Field. We detect excess galaxies around $> 98.4$ per cent of the 850$\mu$m positions in the deep UDS catalogue, giving us the first 850$\mu$m selected sample to have virtually complete optical\slash near-infrared redshift information. Under the reasonable assumption that the redshifts of the excess galaxies are representative of the SMGs themselves, we derive a median SMG redshift of $z = 2.05 \pm 0.03$, with 68 per cent of SMGs residing between $1.07 < z < 3.06$. We find an average of $1.52\pm 0.09$ excess $K$-band galaxies within 12 arc sec of an 850$\mu$m position, with an average stellar mass of $2.2\pm 0.1 \times 10^{10}$ M$_\odot$. While the vast majority of excess galaxies are star-forming, $8.0 \pm 2.1$ per cent have passive rest-frame colours, and are therefore unlikely to be detected at sub-millimetre wavelengths even in deep interferometry. We show that brighter SMGs lie at higher redshifts, and use our SMG redshift distribution -- along with the assumption of a universal far-infrared SED -- to estimate that SMGs contribute around 30 per cent of the cosmic star formation rate density between $0.5 < z < 5.0$.

Abstract:

Large modern surveys require efficient review of data in order to find transient sources such as supernovae, and to distinguish such sources from artefacts and noise. Much effort has been put into the development of automatic algorithms, but surveys still rely on human review of targets. This paper presents an integrated system for the identification of supernovae in data from Pan-STARRS1, combining classifications from volunteers participating in a citizen science project with those from a convolutional neural network. The unique aspect of this work is the deployment, in combination, of both human and machine classifications for near real-time discovery in an astronomical project. We show that the combination of the two methods outperforms either one used individually. This result has important implications for the future development of transient searches, especially in the era of LSST and other large-throughput surveys.