real-time transient searches at the Very Large Array

realfast: real-time transient searches at the Very Large Array

Context

We are developing a set of enhancements to the Jansky Very Large Array (VLA) to enable real-time, commensal surveys for fast radio transients. We use the term realfast to refer to the project and the software.

As the most sensitive cm-wavelength radio interferometer on earth, the VLA will revolutionize this field with its ability to form sensitive images with arcsecond spatial resolution. Interferometric imaging will do pioneering science with a mysterious new class of radio transients known as fast radio bursts. VLA imaging can make the first arcsecond localization of a burst and can detect many events per year. Similarly, the interferometric discovery and localization of other classes of millisecond transient will address topics ranging from the missing baryon problem to (sub)stellar magnetism to the search for exotic binary pulsar systems.

The challenge is that millisecond imaging produces 1 TB of data per hour (in our current standard VLA observing mode) and requires forming many thousands of images per second. So the question is: how do we go from a TB/hour data stream to a simple list of candidates (let us know if you find a transient in that interactive plot of real VLA candidates!)?

We are doing this by developing software and using (and hopefully soon building) computational support at the VLA site. This allows us to process this massive data stream in real time and turn each observation into a fast transient survey, ultimately encompassing thousands of hours per year. Not only does real-time analysis give us the ability to quickly respond to transients, it also makes commensal analysis possible by searching an otherwise inaccessible data stream. Developing this system at the VLA will produce multiplicative improvement to scientific productivity for this already powerful telescope.

Approach

realfast will do this through a series of algorithmic, software, and hardware developments. The principal algorithmic development is presented in rtpipe, a library for building fast transient detection pipelines for radio interferometric data. The largest scale implementation of rtpipe was a 200-hour, 200-TB VLA imaging survey for FRBs. We have also used `rtpipe` to make the first blind interferometric localization of a transient neutron star. For more information on using `rtpipe`, see this blog post.

The current (late 2015) implementation of realfast performs real-time fast imaging to trigger recording of data in a dedicated observation. This system is demonstrating the concepts of "data triage" that are critical to the long-term goal of a commensal system that operates 24/7 and can efficiently excise data around the most promising candidate transients.