J/A+A/666/A1 Strong lensing in UNIONS (Savary+, 2022)
Strong lensing in UNIONS: Towards a pipeline from discovery to modelling.
Savary E., Rojas K., Maus M., Clement B., Courbin F., Gavazzi R.,
Chan J.H.H., Lemon C., Vernardos G., Canameras R., Schuldt S., Suyu S.H.,
Cuillandre J.-C., Fabbro S., Gwyn S., Hudson M.J., Kilbinger M., Scott D.,
Stone C.
<Astron. Astrophys. 666, A1 (2022)>
=2022A&A...666A...1S 2022A&A...666A...1S (SIMBAD/NED BibCode)
ADC_Keywords: Surveys ; Gravitational lensing
Keywords: gravitational lensing: strong - surveys - techniques: image processing
Abstract:
We present a search for galaxy-scale strong gravitational lenses in
the initial 2500 square degrees of the Canada-France Imaging Survey
(CFIS). We design a convolutional neural network (CNN) committee that
we apply on a selection of 2344002 exquisite-seeing r-band images of
color-selected luminous red galaxies (LRGs). Our classification uses a
realistic training set where both the lensing galaxies and the lensed
sources are taken from real data, i.e., the CFIS r-band images
themselves and Hubble Space Telescope (HST). A total of 9460
candidates obtain a score above 0.5 with the CNN committee. After a
visual inspection of the candidates, we find a total of 133 lens
candidates, among which 104 are completely new. The set of false
positives mainly contains ring, spiral and merger galaxies and to a
smaller extent galaxies with nearby companions. We classify 32 of the
lens candidates as secure lenses and 101 as maybe lenses. For the 32
best-quality lenses, we also fit a singular isothermal ellipsoid mass
profile with external shear along with an elliptical Sersic profile
for the lens and source light. This automated modeling step provides
distributions of properties for both sources and lenses which have
Einstein radii in the range 0.5"<θE<2.5". Finally, we
introduce a new lens/source single-band deblending algorithm based on
auto-encoders representation of our candidates. This is the first time
an end-to-end lens-finding and modeling pipeline is assembled
together, in view of future lens searches in single band, as will be
possible with Euclid.
Description:
In this paper we presented the design of an automated pipeline to find
galaxy-scale strong lenses using convolutional neural networks and
applied it to the CFIS wide-field optical imaging survey being carried
out with the 3.6 m CFHT in Hawaii. We used only the deep and sharp
r-band images for which the median seeing is 0.6" down to a 10σ
depth of r=24.6. We used 2500deg2 of CFIS in the present work
since the survey is still ongoing; it is expected to reach a total
area of 5000 deg2 of the northern sky, when completed.
Following the visual inspection, we found 32 objects with striking
lensing features and 101 objects that show strong signs of lensing but
that need further data to confirm (i.e., higher resolution, and deeper
imaging and/or spectroscopy). A by-product of our simulations set is
that we were able to train auto-encoders to learn the lens light and
lensed-source light separately and we then deblended the lens plane
from the source plane for all 133 objects.
We also produced a catalog of contaminants that mimic lensing geometry
(i.e., 238 mergers, 369 ring galaxies, and 961 spiral galaxies).
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
lenses.dat 56 133 Lens candidates obtained after the last stage
of the visual inspection (table 1)
model.dat 102 37 Modeling parameter of the sure lens candidates
(table 2)
spirals.dat 40 957 Objects flagged as spiral galaxies during
visual inspection (table 4)
rings.dat 40 365 Objects flagged as ring galaxies during
visual inspection (table 3)
mergers.dat 40 957 Objects flagged as mergers during visual
inspection (table 5)
--------------------------------------------------------------------------------
Byte-by-byte Description of file: lenses.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 21 A21 --- Name Candidate identifier (UNIONS JHHMMSS+SSMMSS)
23- 27 F5.3 --- CNNscore [0.5/1.0] Output score from the neural network
29- 37 A9 --- Ref Reference(s) for candidates in the
literature (1)
39- 40 I2 h RAh Right ascension (J2000)
42- 43 I2 min RAm Right ascension (J2000)
45- 46 I2 s RAs Right ascension (J2000)
48 A1 --- DE- Declination sign (J2000)
49- 50 I2 deg DEd Declination (J2000)
52- 53 I2 arcmin DEm Declination (J2000)
55- 56 I2 arcsec DEs Declination (J2000)
--------------------------------------------------------------------------------
Note (1): References as follows:
1 = Bolton et al., 2008ApJ...682..964B 2008ApJ...682..964B, Cat. J/ApJ/682/964
2 = Inada et al., 2009AJ....137.4118I 2009AJ....137.4118I
3 = Paraficz et al., 2016A&A...592A..75P 2016A&A...592A..75P, Cat. J/A+A/592/A75
4 = Shu et al., 2016ApJ...833..264S 2016ApJ...833..264S
5 = Shu et al., 2017ApJ...851...48S 2017ApJ...851...48S, Cat. J/ApJ/851/48
6 = Sonnenfeld et al., 2018PASJ...70S..29S 2018PASJ...70S..29S
7 = Cao et al., 2020MNRAS.499.3610C 2020MNRAS.499.3610C
8 = Canameras et al., 2020A&A...644A.163C 2020A&A...644A.163C, Cat. J/A+A/644/A163
9 = Chan et al., 2020A&A...636A..87C 2020A&A...636A..87C
10 = Jaelani et al., 2021MNRAS.502.1487J 2021MNRAS.502.1487J
11 = Huang et al., 2021ApJ...909...27H 2021ApJ...909...27H
12 = Talbot et al., 2021MNRAS.502.4617T 2021MNRAS.502.4617T
--------------------------------------------------------------------------------
Byte-by-byte Description of file: model.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 21 A21 --- Name Candidate identifier (UNIONS JHHMMSS+SSMMSS)
22 A1 --- n_Name [*] Note on Name (1)
24- 28 F5.2 --- rchi2 [0.91/17.3] Reduced chi2
30- 34 F5.3 arcsec RE Einstein radius
36- 40 F5.3 arcsec E_RE Einstein radius error (upper value)
42- 46 F5.3 arcsec e_RE Einstein radius error (lower value)
48- 52 F5.3 --- qm Axis ratio
54- 58 F5.3 --- E_qm Axis ratio error (upper value)
60- 64 F5.3 --- e_qm Axis ratio error (lower value)
66- 68 I3 deg phim Position angle of the lens mass SIE model
70- 71 I2 deg E_phim Position angle of the lens mass SIE model
error (upper value)
73- 74 I2 deg e_phim Position angle of the lens mass SIE model
error (lower value)
76- 80 F5.3 --- gammaext Strength of external shear
82- 86 F5.3 --- E_gammaext Strength of external shear
error (upper value)
88- 92 F5.3 --- e_gammaext Strength of external shear
error (lower value)
94- 96 I3 deg phiext Angle of external shear
98- 99 I2 deg E_phiext Angle of external shear
error (upper value)
101-102 I2 deg e_phiext Angle of external shear
error (lower value)
--------------------------------------------------------------------------------
Note (1): * for the parameters for our best fits, i.e. for a
manually-customized mask, when applicable.
--------------------------------------------------------------------------------
Byte-by-byte Description of file: mergers.dat rings.dat spirals.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 21 A21 --- Name Candidate identifier (UNIONS JHHMMSS+SSMMSS)
23- 24 I2 h RAh Right ascension (J2000)
26- 27 I2 min RAm Right ascension (J2000)
29- 30 I2 s RAs Right ascension (J2000)
32 A1 --- DE- Declination sign (J2000)
33- 34 I2 deg DEd Declination (J2000)
36- 37 I2 arcmin DEm Declination (J2000)
39- 40 I2 arcsec DEs Declination (J2000)
--------------------------------------------------------------------------------
Acknowledgements:
Elodie Savary, elodie.savary(at)epfl.ch
(End) Patricia Vannier [CDS] 22-Jun-2022