Evidence for Two Hot Jupiter Formation Paths

Transcript

1. Evidence For Two Hot Jupiter Formation Paths Ben Nelson CIERA/Northwestern University @exobenelson With: Eric Ford (Penn State) and Fred Rasio (CIERA/NU) arXiv: 1703.09711
2. Disk migration High eccentricity migration Two main camps for explaining these objects Before: After: Hot Jupiter Formation Source: http://jila.colorado.edu/~pja/planet_migration.html
3. In terms of “x”... x ⌘ a aRoche
4. In terms of “x”... x ⌘ a aRoche High eccentricity migration
5. In terms of “x”... x ⌘ a aRoche *not exactly, e.g., Valsecchi & Rasio 2014 x can’t be less than 2!* High eccentricity migration
6. In terms of “x”... x ⌘ a aRoche Disk migration x can’t be less than 2!* *not exactly, e.g., Valsecchi & Rasio 2014 High eccentricity migration
7. In terms of “x”... x ⌘ a aRoche Disk migration x can’t be less than 2!* *not exactly, e.g., Valsecchi & Rasio 2014 **not exactly a prediction from disk migration x can’t be less than 1!** High eccentricity migration
8. High eccentricity migration Lower truncation xl ≈ 2 for RV data x ⌘ a aRoche *not exactly, e.g., Valsecchi & Rasio 2014 Disk migration x can’t be less than 2! x can’t be less than 1! Ford & Rasio 2006
9. Does xl ≈ 2 still hold today? Is it possible to infer a mixture of HJ populations from the current data?
10. The Data Planet samples NASA Exoplanet Archive RV: 52 HAT: 76 WASP: 74 Santerne+ 2016: Kepler: 27 x
11. The Data Planet samples NASA Exoplanet Archive RV :* 52 : 76 ::: 74 Santerne+ 2016: Kepl : 27 x *Credit: https://twitter.com/erikanesvold/status/846414211903344640
12. The Model: 1-component j = 1, · · · , n xl ˆ Pj ˆ Rj ˆ Rj Rj xu ij xj ˆ Mj ˆ Mj Pj Mj ˆ Pj made with daft daft-pgm.org
13. The Model: 1-component j = 1, · · · , n xl ˆ Pj ˆ Rj ˆ Rj Rj xu ij xj ˆ Mj ˆ Mj Pj Mj ˆ Pj population-level parameters made with daft daft-pgm.org
14. The Model: 1-component j = 1, · · · , n xl ˆ Pj ˆ Rj ˆ Rj Rj xu ij xj ˆ Mj ˆ Mj Pj Mj ˆ Pj individual parameters probabilistic graphical model with daft: daft-pgm.org made with daft daft-pgm.org
15. The Model: 1-component j = 1, · · · , n xl ˆ Pj ˆ Rj ˆ Rj Rj xu ij xj ˆ Mj ˆ Mj Pj Mj ˆ Pj “observational” data made with daft daft-pgm.org
16. j = 1, · · · , n k = 1, 2 k x1,k ˆ Mj ˆ Pj fk ˆ Rj Rj xu ij xj Pj ˆ Mj ˆ Rj Mj ˆ Pj The Model: 2-component mixture made with daft daft-pgm.org
17. Stan Sampling high dimensions with
18. Stan Sampling high dimensions with
19. Stan Sampling high dimensions with
20. Hamiltonian Monte Carlo arXiv: 1701.02434 Stan Sampling high dimensions with
21. RV+Kepler data do not show evidence for multiple populations p(d|M 2comp ) p(d|M 1comp ) Bayes Factor ≈ 0.1 = * *github.com/benelson/FML
22. Lower truncation xl ≈ 2 is still there
23. HAT+WASP data show evidence for multiple populations ≈ 1021 = p(d|M 2comp ) p(d|M 1comp ) Bayes Factor* *github.com/benelson/FML
24. x p(x) xl,1 xl,2 xu 1 1 2 1 Locations of inner edges around where the theories predict
25. x p(x) xl,1 xl,2 xu 1 1 2 1 Most HJs reside in component consistent with eccentric migration
26. Correlations with host star properties? x
27. Within the limitations of our chosen models... Multiple HJ populations can be inferred from current data. RV+Kepler data are well explained as a single population with xl ≈ 2. For HAT+WASP data... 85% consistent with high-e migration history 15% consistent with disk migration history
28. Within the limitations of our chosen models... Multiple HJ populations can be inferred from current data. RV+Kepler data are well explained as a single population with xl ≈ 2. For HAT+WASP data... 85% consistent with high-e migration history 15% consistent with disk migration history Caveats including but not limited to: 1. NASA Exoplanet Archive query missed some planets due to filtering issues; 2. similar analysis with exoplanets.org data show moderately different constraints on some pop-level parameters; 3. smaller sample size for RV and Kepler, more data could reveal 1<x<2 planets in these datasets; 4. we only include data on x, nothing on obliquity, additional companions, etc.; 5. unclear if HAT and WASP have significant observing biases affecting shape of x distribution; 6. large asymmetries in planet radius estimates not accurately modeled with Gaussian; 7. hard edge most sensitive to planet with smallest x, doesn’t offer enough flexibility for outliers * *
29. Full analysis publicly available github.com/benelson/hjs_with_stan - reproducible results - propose alternative Stan models - adopt Stan models for your own research problem
30. Expected to find ~10,000 Hot Jupiters (Sullivan+ 2015)
31. Recovering the input properties of synthetic populations
32. Recovering the input properties of synthetic populations
33. Within the limitation of our chosen models, multiple HJ populations can be inferred from current data. • RV+Kepler data are well explained as a single population with xl ≈ 2. • For HAT+WASP data... - 85% consistent with high-e migration history - 15% consistent with disk migration history Paper at arXiv:1703.09711 Full analysis available at github.com/benelson/hjs_with_stan Summary