The song and the cacophony; searching LSST for eclipsing white dwarf binaries
Lead Research Organisation:
University of Sheffield
Abstract
Double white dwarf binaries are an elusive population of our Galaxy and key to understanding a range of topics in binary evolution including the population of Type Ia supernova and R Corona Borealis stars. They also dominate the population of detectable gravitational-wave sources, with more than ten thousand potential sources in the Milky Way. This opens up the study of general relativity, tidal physics in degenerate matter and Galactic structure. They will also represent a challenge for the upcoming LISA mission, as a large background source of gravitational radiation. Eclipsing double white dwarf binaries are particularly important, as they allow detailed parameter studies of the components.
Despite their importance, the faint optical emission and short timescale variability from eclipsing double white dwarfs makes them a challenge to find in significant numbers. The most promising method is to search for their periodic variability in all-sky synoptic surveys. Recent exploitation of light curves from ZTF has increased the number of known eclipsing double white dwarf binaries with short periods from two to nine, but larger samples require a deeper survey.
LSST represents both an unprecedented opportunity and an enormous challenge for finding double white dwarf binaries. Simulations predict that LSST can detect of order one thousand double white dwarf binaries, but identifying them will require searching billions of objects for periodic variability.
This is not computationally feasible using existing approaches to period search as tools such as the Lomb-Scargle periodogram are too slow, even when accelerated on GPUs. In this proposal we will develop a new, modern approach for period-searches by building a machine-learning powered emulator for existing period search algorithms. This emulator can be used by all astronomers looking for periodic variability to compute their periodicity statistic of choice in a computationally cheap way. We will make this tool publicly available, and deploy it to deliver a step-change in our understanding of the Milky Way's double white dwarf binary population.
Despite their importance, the faint optical emission and short timescale variability from eclipsing double white dwarfs makes them a challenge to find in significant numbers. The most promising method is to search for their periodic variability in all-sky synoptic surveys. Recent exploitation of light curves from ZTF has increased the number of known eclipsing double white dwarf binaries with short periods from two to nine, but larger samples require a deeper survey.
LSST represents both an unprecedented opportunity and an enormous challenge for finding double white dwarf binaries. Simulations predict that LSST can detect of order one thousand double white dwarf binaries, but identifying them will require searching billions of objects for periodic variability.
This is not computationally feasible using existing approaches to period search as tools such as the Lomb-Scargle periodogram are too slow, even when accelerated on GPUs. In this proposal we will develop a new, modern approach for period-searches by building a machine-learning powered emulator for existing period search algorithms. This emulator can be used by all astronomers looking for periodic variability to compute their periodicity statistic of choice in a computationally cheap way. We will make this tool publicly available, and deploy it to deliver a step-change in our understanding of the Milky Way's double white dwarf binary population.