Chitah: STRONG-GRAVITATIONAL-LENS HUNTER IN IMAGING SURVEYS

Abstract

Strong gravitationally lensed quasars provide powerful means to study galaxy evolution and cosmology. Current and upcoming imaging surveys will contain thousands of new lensed quasars, augmenting the existing sample by at least two orders of magnitude. To find such lens systems, we built a robot, Chitah, that hunts for lensed quasars by modeling the configuration of the multiple quasar images. Specifically, given an image of an object that might be a lensed quasar, Chitah first disentangles the light from the supposed lens galaxy and the light from the multiple quasar images based on color information. A simple rule is designed to categorize the given object as a potential four-image (quad) or two-image (double) lensed quasar system. The configuration of the identified quasar images is subsequently modeled to classify whether the object is a lensed quasar system. We test the performance of Chitah using simulated lens systems based on the Canada-France-Hawaii Telescope Legacy Survey. For bright quads with large image separations (with Einstein radius r ein > 1. ′′ 1) simulated using Gaussian point-spread functions, a high true-positive rate (TPR) of ∼ /90% and a low false-positive rate of ∼3% show that this is a promising approach to search for new lens systems. We obtain high TPR for lens systems with r ein 0.5 ′′ , so the performance of Chitah is set by the seeing. We further feed a known gravitational lens system, COSMOS 5921+0638, to Chitah, and demonstrate that Chitah is able to classify this real gravitational lens system successfully. Our newly built Chitah is omnivorous and can hunt in any ground-based imaging surveys.

Discovery of the Lensed Quasar System DES J0408-5354

“…These other search methods are described in more detail in Agnello et al (2017). We provide here a brief summary: (1) the Gaussian Mixture Model (GMM) method of Ostrovski et al (2017),w h i c h uses supervised machine learning in a five-dimensional optical plus infrared color space and identified DES J0408-5354 as a candidate pair (D+G1 not found separately in J band due to blending noted above); (2) CHITAH (Chan et al 2015), which uses pixel-based automatic recognition on grizY cutout images and identified the system as a candidate pair (not flagged as a quad because the fourth image G2 is too red);and(3) the Artificial Neural Network method of Agnello et al (2015aAgnello et al ( , 2015b, which uses griz and W W 1 2 magnitudes and identified DES J0408-5354 as a candidate extended quasar. Figure 2 shows how the system was flagged by the GMM method as a candidate, due to the quasar-like colors of its components.…”

Discovery of two gravitationally lensed quasars in the Dark Energy Survey

“…Fortunately, the DES image quality (median seeing ∼ 0.9 ) is better than that of SDSS, so one can rely on more accurate morphological information for candidate selection. New techniques have been developed in order to address this challenge (Agnello et al 2015;Chan et al 2015).…”

Image Simulations for Strong and Weak Gravitational Lensing

“…Image simulations of strong lensing systems have been used to predict that current and future wide-field galaxy surveys (e.g, DES, HSC survey, KiDS, LSST, Euclid, and WFIRST ) will produce several to hundreds of thousands of galaxy-galaxy strong lensing systems (Collett 2015). Many efforts have recently focused on employing techniques from computer vision and machine learning to go beyond traditional approaches such as visual searches of "blue" arcs near "red" galaxies (Diehl et al 2017), goodness of fit examinations after fitting a model to all candidates (Marshall et al 2009;Chan et al 2015), and public science challenges to discover new strong lensing systems in the large datasets. Neural networks have demonstrated to be able to distinguish between simulated lenses and non-lenses (Lanusse et al 2017;Hezaveh et al 2017).…”