Exploring the Fourth Dimension: Enhancing the Discovery Power of Transient Surveys with an Effective Classification/Identification System

Time-domain astronomy is the new frontier of discovery. We now have the technical capabilities to monitor the night sky in real time and explore the fourth dimension: TIME. This “observational exercise” has led to the discovery of new types of astrophysical phenomena that challenge current notions of physics. Recent major discoveries include the unexpected finding of a class of rare but extremely luminous stellar explosions of unknown origin; the finding of very short burst of radio waves from extragalactic space; the discovery of numerous bursts of radiation originating from stellar disruptions by super-massive black holes located at the center of galaxies, and, more recently, the landmark discovery of light from a gravitational-wave source. These discoveries are revolutionizing our understanding of physics in the regime of extreme gravity -which cannot be probed otherwise-, and are transforming our notions of stellar evolution, -which directly impacts astronomy at large, e.g. our understanding of the star formation history and chemical enrichment of the Universe-. Transient astronomical surveys at all wavelengths are thus effectively opening a new window of investigation on our universe. Current and near-future surveys like the Young Stellar Experiment (YSE, led by USCS, 2018+), the Zwicky Transient Facility (ZTF, led by Caltech, 2018+) and the Large Synoptic Survey Telescope (LSST, 2022+) will take this effort to the next level and will generate an unprecedented data stream in terms of volume, variety, and velocity, thus effectively offering an unparalleled opportunity for ground-breaking discoveries. Transient surveys will find millions of new astronomical sources in the sky each night of observation. For each of these new sources, an “alert” will be issued with minimal information about the source, which includes coordinates and apparent brightness in a given optical filter. In this context, our ability to make historic and major discoveries in this upcoming era of data-driven large-scale time-domain astronomical surveys is limited by a two-pronged challenge: (i) it will be necessary to identify specific targets of interest amidst millions of alerts; (ii) we will need very fast identification of notable alerts, ideally within minutes of detection, to enable prompt follow-up observations. Follow-up observations consist of spectroscopic data and/or observations across the electromagnetic spectrum -including X-ray and radio data- and represent the only way to turn “detections” (i.e. the knowledge that something new appeared in the sky), into “discoveries” (which imply a characterization of the properties of the source and a full understanding of its physics and intrinsic nature). Finding a solution to (i)+(ii) is urgent: ZTF is now online and LSST will start operations in the next few years. Our project specifically addresses (i)+(ii), with the overarching goal to provide the entire astronomical community with a tool that enables prompt and effective classification of transient alerts [Fig. 1]. We emphasize that this is the first necessary step to realize the tremendous scientific potential of time domain astronomy. Our end product is a probabilistic statement about the intrinsic nature of each alert that can be distributed promptly, i.e. a “number” that quantify our knowledge/ignorance of the nature of the alert at the time of detection. While this output is of scientific relevance on its own, the maximum benefit to time-domain astronomy can only be achieved through deep integration with the other components of t