New UX Ori ri type pe candi ndida dates de detect cted d us - PowerPoint PPT Presentation

New UX Ori ri type pe candi ndida dates de detect cted d us using ng Gaia DR2 and nd Ma Machi chine ne Learni rning ng Miguel Vioque University of Leeds R. D. Oudmaijer (University of Leeds, UK), M. Schreiner (Desupervised, Denmark), D. Baines (ESAC, Spain), and R. Pérez-Martínez (Isdefe, Spain) The UX Ori type stars and related topics, 1st of October, 2019

Hi High gh-Ma Mass Star r Form rmation Herbig Ae/Be stars Spectral types ~ F5 to B0 Mass: 2 − 10𝑁 ⨀ Intermediate-mass T-Tauri stars T-Tauri stars Alecian, et al. (2013), Villebrun, et al. (2019)

Hi High gh-Ma Mass Star r Form rmation Around 260 known to date Herbig Ae/Be stars Spectral types ~ F5 to B0 Mass: 2 − 10𝑁 ⨀ Intermediate-mass T-Tauri stars T-Tauri stars Alecian, et al. (2013), Villebrun, et al. (2019)

Hi High gh-Ma Mass Star r Form rmation

Hi High gh-Ma Mass Star r Form rmation Gaia DR2 All known Herbig Ae/Be stars Vioque, et al. (2018)

Hi High gh-Ma Mass Star r Form rmation Gaia DR2 All known Herbig Ae/Be stars Vioque, et al. (2018)

Gaia vari riabi bility ) ~ 𝜏 𝐺 , 𝑂 ./0 𝑊 𝐺 , Deason et al. 2017 Vioque et al. 2018

Gaia vari riabi bility ) ~ 𝜏 𝐺 , 𝑂 ./0 𝑊 𝐺 , Deason et al. 2017 All known 𝑊 ) > 2 UXORs in the sample (17) Vioque et al. 2018 With variabilities larger than 0.5mag …

Gaia vari riabi bility ) ~ 𝜏 𝐺 , 𝑂 ./0 𝑊 𝐺 , Deason et al. 2017 Proposed 31 new UX Ori candidates All known 𝑊 ) > 2 UXORs in the sample (17) Vioque et al. 2018 With variabilities larger than 0.5mag …

Looking for new Pre-Main Sequence (PMS) objects in Gaia! Main characteristics of PMS objects: • Infrared excesses • H 𝛽 emission • Photometric variability

Looking for new Pre-Main Sequence (PMS) objects in Gaia! Main characteristics of PMS objects: • Infrared excesses • H 𝛽 emission • Photometric variability High mass PMS objects (Herbig Be stars) are very similar to Classical Be stars ... and supergiants, B[e] stars, …

Looking for new Pre-Main Sequence (PMS) objects in Gaia! Perform an homogeneous selection, distance and Main characteristics of PMS objects: • Infrared excesses position independent! • H 𝛽 emission • Photometric variability High mass PMS objects (Herbig Be stars) are very similar to Classical Be stars ... and supergiants, B[e] stars, …

Neur ural Network rk Algorithm is trained with known labeled data After generalizing: Before training: Each category gets • Training Set • a probability Set of characteristics • Efficiency of the • Set of categories • algorithm The best architecture is selected

Neur ural Network rk Selection of the characteristics : AllWISE (WISE+2MASS) 𝑿𝟐, 𝑿𝟑, 𝑿𝟒, 𝑿𝟓 • Infrared excesses 𝑲, 𝑰, 𝑳 𝒕 IPHAS VPHAS+ 𝒔 − 𝑰 𝜷 • H 𝛽 emission Gaia 2 variability indicators • Photometric variability 𝑪 𝒒 , 𝑯, 𝑺 𝒒

Neur ural Network rk Create all possible colours Distance independent! Selection of the characteristics : AllWISE (WISE+2MASS) 𝑿𝟐, 𝑿𝟑, 𝑿𝟒, 𝑿𝟓 • Infrared excesses 𝑲, 𝑰, 𝑳 𝒕 IPHAS VPHAS+ 𝒔 − 𝑰 𝜷 • H 𝛽 emission Gaia 2 variability indicators • Photometric variability 𝑪 𝒒 , 𝑯, 𝑺 𝒒

Selection of the categories : PMS category Classical Be category Other sources

Selection of the Training Set : AllWISE Gaia IPHAS VPHAS+ = 4,151,538 + + sources PMS category Classical Be category Other sources

Selection of the Training Set : AllWISE Gaia IPHAS VPHAS+ = 4,151,538 + + sources 848 Pre-Main Sequence • PMS category objects ( 163 Herbig Ae/Be) Classical Be 775 Classical Be stars • category 471,111 random sources • Other sources

Training the Neural Network

Trained Neural Network AllWISE Gaia = 4,151,538 IPHAS VPHAS+ + + sources

Proba babi bility Ma Map

Proba babi bility Ma Map 636 Classical Be candidates 1266 either 8452 PMS candidates 4,140,629 other

Proba babi bility Ma Map 636 Classical Evaluation on Test Set Be candidates PMS Completeness 𝟖𝟗. 𝟗 ± 𝟐. 𝟓% Classical Be 1291 either Completeness 𝟗𝟔. 𝟔 ± 𝟐. 𝟑% 8452 PMS candidates 4,140,629 other

Gaia HR di diagram PMS candidates > 50% PMS candidates > 65% ¡ 4 ¡ 4 ¡ 2 ¡ 2 0 0 2 2 M G [mag] M G [mag] 4 4 6 6 8 8 10 10 12 12 14 14 PMS candidates > 80% PMS candidates > 95% ¡ 1 ¡ 1 0 1 2 3 4 0 1 2 3 4 ¡ 4 ¡ 4 B p ¡ R p B p ¡ R p ¡ 2 ¡ 2 0 0 2 2 M G [mag] M G [mag] 4 4 6 6 8 8 10 10 12 12 14 14 ¡ 1 ¡ 1 0 1 2 3 4 0 1 2 3 4 B p ¡ R p B p ¡ R p

Co Coordi dina nates

Co Coordi dina nates

Co Coordi dina nates

UX Ori ri candi ndida dates Vioque et al. 2018

UX Ori ri candi ndida dates Proposed 31 new UX Ori candidates Vioque et al. 2018

UX Ori ri candi ndida dates 350 300 250 Gaia Variability 200 3436 UX Ori 150 candidates 100 50 0 : 5 0 : 6 0 : 7 0 : 8 0 : 9 1 : 0 Probability PMS Vioque et al. in prep

UX Ori ri candi ndida dates 350 300 250 Gaia Variability ~40% of the PMS candidates 200 3436 UX Ori 150 candidates 100 50 0 : 5 0 : 6 0 : 7 0 : 8 0 : 9 1 : 0 Probability PMS Vioque et al. in prep

Ca Caveats PMS Candidates H 𝛽 emission Main Sequence Reddening

Ca Caveats PMS Candidates H 𝛽 emission Main Sequence Reddening

Caveats Ca Planetary Nebulae! PMS Candidates H 𝛽 emission Main Sequence Reddening

Fut Futur ure work rk Past and future observations + + 2.2m Calar Alto INT NTT Populate HR diagram 𝟑𝟕𝟏 objects + ~𝟒𝟏𝟏𝟏 objects ¡ 6 ¡ 6 0 : 95 ¡ 4 ¡ 4 0 : 90 ¡ 2 ¡ 2 Probability PMS 0 : 85 0 0 M G [mag] M G [mag] 0 : 80 2 2 0 : 75 4 4 0 : 70 6 6 0 : 65 8 8 0 : 60 10 10 0 : 55 12 12 ¡ 1 ¡ 1 0 1 2 3 4 0 1 2 3 4 B p ¡ R p B p ¡ R p

Resul ults We retrieve 8452 new PMS • Completeness candidates. 3131 potential 𝟖𝟗. 𝟗 ± 𝟐. 𝟓% Herbig Ae/Be stars. Gaia 2204517656901678848 𝐻 = 14.0mag , 𝑒 = 940 pc We retrieve 3436 new UX Ori type stars candidates IDS/INT 𝐼 M line spectra We retrieve 636 new Classical Completeness • Be stars candidates. 𝟗𝟔. 𝟔 ± 𝟐. 𝟑%