J/A+A/704/A150 ExoDNN v1 candidates (Abreu+, 2025)
ExoDNN: Boosting exoplanet detection with artificial intelligence.
Application to Gaia Data Release 3.
Abreu A., Lillo-Box J., Perez-Garcia A.M., Sahlmann J., de Bruine J.H.,
Cifuentes C.
<Astron. Astrophys. 704, A150 (2025)>
=2025A&A...704A.150A 2025A&A...704A.150A (SIMBAD/NED BibCode)
ADC_Keywords: Stars, double and multiple ; Optical
Keywords: astrometry - planets and satellites: detection - binaries: general
Abstract:
Transit and radial velocity (RV) techniques are the dominant methods
for exoplanet detection, while astrometric exoplanet detections have
been very limited thus far. Gaia has the potential to radically change
this picture, enabling astrometric detections of substellar companions
at scale that would allow us to complement the picture of exoplanet
architectures given by transit and RV methods.
Our primary objective in this study is to enhance the current
statistics of substellar companions, particularly within regions of
the orbital period-mass parameter space that remain poorly constrained
by RV and transit detection methods.
Using supervised learning, we trained a deep neural network (DNN) to
recognise the characteristic distribution of the fit quality
statistics corresponding to a Gaia Data Release 3 (DR3) astrometric
solution for a non-single star. We created a deep learning model,
ExoDNN, which predicts the probability of a DR3 source to host
unresolved companions.
Applying the predictive capability of ExoDNN to a volume-limited
sample (d<100pc) of F, G, K, and M stars from Gaia DR3, we have
produced a list of 7414 candidate stars hosting companions. The
stellar properties of these candidates, such as their mass and
metallicity, are similar to those of the Gaia DR3 non-single-star
sample. We also identified synergies with future observatories, such
as PLATO, and we propose a follow-up strategy with the intention of
investigating the most promising candidates among those samples.
Description:
Using supervised learning, we have trained a deep neural network to
recognize the characteristic distribution of the fit quality
statistics corresponding to a DR3 astrometric solution for a
non-single star. We have generated a deep learning model, ExoDNN, that
predicts the probability of a DR3 source to host unresolved
companions.When applied to a sample of stars from the Gaia DR3
catalogue within 100pc, ExoDNN generated a list of 7414 new candidate
stars to host unresolved companions. The number of candidates
detected, is comparable in order of magnitude to predictions from
earlier studies (e.g. Casertano et al. 2008A&A...482..699C 2008A&A...482..699C ; Perryman
et al. 2014ApJ...797...14P 2014ApJ...797...14P). A false positive rate of ∼1.2% is
expected from the current version of ExoDNN, and should be taken into
account when assessing the proposed candidates by performing
additional scrutiny on a per-source basis.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
exodnnv1.dat 386 7414 File with the 7414 Gaia DR3 candidate stars to
host unresolved companions produced by ExoDNN
--------------------------------------------------------------------------------
See also:
I/355 : Gaia DR3 Part 1. Main source (Gaia Collaboration, 2022)
Byte-by-byte Description of file: exodnnv1.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 31 A31 --- Name Common Source Name
33- 40 A8 --- --- [Gaia DR3]
42- 60 I19 --- GaiaDR3 Gaia DR3 unique source designation
(source_id)
62- 69 F8.4 deg RAdeg Right ascension (ICRS) at Ep=2016.0
71- 76 F6.4 deg e_RAdeg Standard error of right ascension
78- 85 F8.4 deg DEdeg Declination (ICRS) at Ep=2016.0
87- 92 F6.4 deg e_DEdeg Standard error of declination
94-103 F10.4 mas/yr pmRA Proper motion in right ascension direction
105-110 F6.4 mas/yr e_pmRA Standard error of proper motion in
right ascension direction
112-121 F10.4 mas/yr pmDE Proper motion in declination direction
123-128 F6.4 mas/yr e_pmDE Standard error of proper motion in
declination direction
130-137 F8.4 mas Plx Absolute barycentric stellar parallax of
the source at the reference epoch 2016
139-144 F6.4 mas e_Plx Standard error of parallax
146-152 F7.4 --- RUWE Renormalised unit weight error
154-159 F6.4 mas epsi Excess noise of the source
161-172 F12.4 --- chi2AL AL chi-square value
174-181 F8.4 --- gofAL Goodness of fit statistic of model wrt
along-scan observations
183-186 I4 --- NAL Total number of observations in the
along-scan (AL) direction
188-200 F13.4 e-/s FG G-band mean flux
202-211 F10.4 e-/s e_FG Error on G-band mean flux
213-222 F10.4 --- RFG G-band mean flux over its error
224-230 F7.4 mag Gmag G-band mean magnitude
232-244 F13.4 e-/s FBP BP-band mean flux
246-255 F10.4 e-/s e_FBP Error on BP-band mean flux
257-265 F9.4 --- RFBP BP-band mean flux over its error
267-273 F7.4 mag BPmag Integrated BP mean magnitude
275-287 F13.4 e-/s FRP RP-band mean flux
289-299 F11.4 e-/s e_FRP Error on RP-band mean flux
301-310 F10.4 --- RFRP RP-band mean flux over its error
312-318 F7.4 mag RPmag Integrated RP mean magnitude
320-325 F6.4 mag E(BP/RP) BP/RP excess factor
327-332 F6.4 mag BP-RP BP-RP colour
334-339 F6.4 mag BP-G BP-G colour
341-346 F6.4 mag G-RP G-RP colour
348-356 F9.4 km/s RV Radial velocity
358-364 F7.4 km/s e_RV Radial velocity error
366-372 F7.4 mag GRVSmag ?=- Integrated Grvs magnitude
374-379 F6.4 mag e_GRVSmag ?=- Grvs magnitude uncertainty
381-386 F6.4 --- PredProb1 Model predicted probability of hosting
unresolved companion
--------------------------------------------------------------------------------
Acknowledgements:
Asier Abreu, asierabreu(at)gmail.com
(End) Patricia Vannier [CDS] 06-Oct-2025