One of the major challenges for modern supernova surveys is identifying the galaxy that hosted each explosion. Is there an accurate and efficient way to do this that avoids investing significant human resources?
Why Identify Hosts?Supernovae are a critical tool for making cosmological predictions that help us to understand our universe. But supernova cosmology relies on accurately identifying the properties of the supernovae — including their redshifts. Since spectroscopic followup of supernova detections often isn’t possible, we rely on observations of the supernova host galaxies to obtain redshifts.
But how do we identify which galaxy hosted a supernova? This seems like a simple problem, but there are many complicating factors — a seemingly nearby galaxy could be a distant background galaxy, for instance, or a supernova’s host could be too faint to spot.
Turning to Automation
Before the era of large supernovae surveys, searching for host galaxies was done primarily by visual inspection. But current projects like the Dark Energy Survey’s Supernova Program is finding supernovae by the thousands, and the upcoming Large Synoptic Survey Telescope will likely discover hundreds of thousands. Visual inspection will not be possible in the face of this volume of data — so an accurate and efficient automated method is clearly needed!
To this end, a team of scientists led by Ravi Gupta (Argonne National Laboratory) has recently developed a new automated algorithm for matching supernovae to their host galaxies. Their work builds on currently existing algorithms and makes use of information about the nearby galaxies, accounts for the uncertainty of the match, and even includes a machine learning component to improve the matching accuracy.
Gupta and collaborators test their matching algorithm on catalogs of galaxies and simulated supernova events to quantify how well the algorithm is able to accurately recover the true hosts.
Successful MatchingThe authors find that when the basic algorithm is run on catalog data, it matches supernovae to their hosts with 91% accuracy. Including the machine learning component, which is run after the initial matching algorithm, improves the accuracy of the matching to 97%.
The encouraging results of this work — which was intended as a proof of concept — suggest that methods similar to this could prove very practical for tackling future survey data. And the method explored here has use beyond matching just supernovae to their host galaxies: it could also be applied to other extragalactic transients, such as gamma-ray bursts, tidal disruption events, or electromagnetic counterparts to gravitational-wave detections.
Ravi R. Gupta et al 2016 AJ 152 154. doi:10.3847/0004-6256/152/6/154