A whitened matched-filter statistic for detecting single exoplanet transits in survey photometry

A whitened matched-filter statistic for detecting single exoplanet transits in survey photometry

Exoplanet detection

The TESS mission has discovered many long-period exoplanets which transit only once. These planets represent an understudied population of temperate planets and recovering their orbital periods would provide insightful tests for models of planet formation and migration. With significant follow-up of such planets from both space- and ground-based observatories, there is a need to efficiently detect epochs which are well matched to known transit shapes in photometric datasets. This report describes the calibration of the template-matching search adopted in the BRUCE package, a Python library for modelling and fitting transiting planets and eclipsing binaries. The likelihood-ratio statistic returned by the scan is standardised into a matched-filter signal-to-noise ratio (S/N), z, which is exactly Gaussian under white noise, before being whitened by its own median absolute deviation over each scan. Six noise conditions, each with 2,500 realisations, were used to measure the false-alarm rate at local probabilities of 10^(-3) and 10^(-4). Under quasi-periodic spot modulation and granulation, the raw statistic produced false alarms at 230 and 103 times the nominal rate at 10^(-3) and up to 1,900 times at 10^(-4). The whitened statistic held every condition to within factors of 1.3 and 1.9, respectively. To test the sensitivity of the search, we performed an injection-recovery test on a grid of 51,200 synthetic TESS light curves: recovery was found to follow lines of constant single-transit S/N and the false-peak rate to scale inversely with template width. We conclude that this whitened matched-filter S/N statistic is the correct choice for the pipeline, as a single threshold is valid under both white and correlated noise, allowing further TESS single-transit candidates to be detected with a known false-alarm probability.

Exoplanet, TESS, NGTS