SELFI: an object-based, Bayesian method for faint emission line source detection in MUSE deep field data cubes

Abstract

We present SELFI, the Source Emission Line FInder, a new Bayesian method optimized for detection of faint galaxies in Multi Unit Spectroscopic Explorer (MUSE) deep fields. MUSE is the new panoramic integral field spectrograph at the Very Large Telescope (VLT) that has unique capabilities for spectroscopic investigation of the deep sky. It has provided data cubes with 324 million voxels over a single 1 arcmin 2 field of view. To address the challenge of faint-galaxy detection in these large data cubes, we developed a new method that processes 3D data either for modeling or for estimation and extraction of source configurations. This objectbased approach yields a natural sparse representation of the sources in massive data fields, such as MUSE data cubes. In the Bayesian framework, the parameters that describe the observed sources are considered random variables. The Bayesian model leads to a general and robust algorithm where the parameters are estimated in a fully data-driven way. This detection algorithm was applied to the MUSE observation of Hubble Deep Field-South. With 27 h total integration time, these observations provide a catalog of 189 sources of various categories and with secured redshift. The algorithm retrieved 91% of the galaxies with only 9% false detection. This method also allowed the discovery of three new Lyα emitters and one [OII] emitter, all without any Hubble Space Telescope counterpart. We analyzed the reasons for failure for some targets, and found that the most important limitation of the method is when faint sources are located in the vicinity of bright spatially resolved galaxies that cannot be approximated by the Sérsic elliptical profile.

The MUSE Hubble Ultra Deep Field Survey

“…Several tools have already been developed to perform blind searches of faint emitters in MUSE datacubes, such as: MUSELET, a SExtractor based method available in MPDAF 11 , LSDCAT, a matching filter method , SELFI, a Bayesian method (Meillier et al 2016) and CubExtractor (Cantalupo, in prep. ), a three-dimensional automatic extraction software based on connecting-labeling-component algorithm (used, e.g., in Borisova et al 2016;.…”

Detecting Weak Spectral Lines in Interferometric Data through Matched Filtering

“…Cross-correlating a noisy signal with this filter maximizes the output S/N. This approach is used extensively in digital signal processing; prominent examples include RADAR (e.g., Woodward 1953;Cumming & Wong 2005, and references therein), source detection in imaging surveys (e.g., Bertin & Arnouts 1996;Bertin 2001;Meillier et al 2016;Herenz & Wisotzki 2017;Zackay & Ofek 2017), gravitational wave detection (e.g., Owen & Sathyaprakash 1999;Schutz 1999;Abbott et al 2016), and exoplanet detection through direct imaging (Ruffio et al 2017).…”

A 16 deg2 survey of emission-line galaxies at z < 1.5 in HSC-SSP Public Data Release 1

“…However, target selection for spectroscopy can cause biases with the observed galaxy sample. Rather than slit or fiber spectroscopy, an integral field unit spectrograph such as VLT/MUSE (Bacon et al 2015;Meillier et al 2016;Bina et al 2016;Swinbank et al 2017) or grism such as HST/WFC3 (e.g., Atek et al 2010;van Dokkum et al 2011;Straughn et al 2011;van der Wel et al 2011;Brammer et al 2012;Colbert et al 2013;Pirzkal et al 2013;Mehta et al 2015;Morris et al 2015;Momcheva et al 2016) can overcome selection biases by allowing us to get spectrum of all galaxies over the field surveyed. One weakness is the small field of view (FoV).…”