Navigation :

Planet-Star Connections in the Era of TESS and Gaia

May 20-24, 2019

Kavli Institute for Theoretical Physics, UCSB, Santa Barbara, CA

All talks recorded and slides posted here

People going from Penn State

Angie Wolfgang
Zhao Guo
Dan Stevens

Monday, May 20:

Vincent Van Eylen - What astereoseismology can do for planets. His radius gap paper based on asteroseismology. Also eccentricity distribution for transiting planets using asteroseismology to better constrain the mean stellar density, which factors into the transit duration (single transiting systems have higher eccentricity). Also obliquity measurements enabled by asteroseismology. Future for exoplanets and asteroseismology: look at more evolved stars: currently few planets detected around subgiant stars; most have asteroseismology.

Angela Santos - asteroseismic signatures of magnetic activity in solar-type stars. Looking at frequency shifts over hundreds of days (magnetic activity produces cyclic variations in the oscillation properties - Santos 2018: frequency shifts are anti-correlated with lnS). Presented results for individual stars; also, see strong relation between frequency shift as a function of age (less strong with radius).

Jorgen Christensen-Dalsgaard - asteroseismology overview. Key point: stellar modeling is involved in order to go from observed frequency of oscillations of mass, radius and age. So if our stellar models are wrong, the asteroseismic M,R are wrong. Microophysics in stellar models: opacity, diffusion and settling (these are most uncertain and is topic of his talk), equation of state, nuclear reaction rates. A recent revision of solar composition changed the the solution of the sound speed to be off by up to 3% - maybe a problem with opacity, particularly helium - find different He abundance as a function of layer in star for different Kepler stars, which implies different diffusion rates for helium. Differences produced by mixing? Can constrain convective overshoot distance with asteroseismology: increases as a function of mass up to limit of 0.15 times scale height.

Travis Berger - same talk as Kepler/K2 SciCon

James Owen - standard photoevaporation talk + Owen & Wu, 2018 results on metallicity difference between “born rocky” and photoevaporated planets (division at P = 25 days).

Hans Kjeldsen - low SNR detections of p-modes in TESS data. Signal is affected by Teff^¹⁄₂, observation wavelength (1/lambda), Nyquist sampling, and coherence of the signal (~ Teff^-³⁄₂). Kepler-10 is example of low asteroseismic SNR (revisited in 2014 - still got 2.4% error in mass, 0.8% error in radius, 13.5% error in age because of long baseline). For many stars observed with TESS, nu_max will not be observable with the 2-minute cadence. There is a relation between delta-nu and nu_max that can help; also well differentiated evolution in delta-nu_0 (the standard delta-nu, which is between consecutive radial modes) and delta-nu_0,2 (the difference between the l=0 and l=2 modes within one radial mode, I think?). Can combine ground-based and TESS data to improve SNR.

Tim Bedding - next frontiers for asteroseismology: 1) validating the asteroseismic scaling relations; 2) rotation/angular momentum transport; 3) asteroseismology of A-type stars (particularly delta-Scuti pulsators; just off the main sequence). Scaling relations: delta-nu ~ rho^¹⁄₂; nu_max ~ g*Teff^-¹⁄₂. delta-nu IS APPROXIMATE but departures are well understood from theoretical models by 2-10% (White et al. 2011). nu-max is more problematic. Physics is less understood here; depends on reflection off boundary between convective and radiative zone in the star (characterized by the acoustic cut-off frequency), and the assumption that nu-max is directly proportional to this cut-off frequency. So far this relation works … but needs to be tested. Example of cause for concern: width of oscillation envelope depends on mass of star … does it do so symmetrically? Rotation - observable through splitting in l=1 mode (m=-1,0,1); height of m=1,-1 modes depends on inclination. Gamma doradus stars - can measure the rotation of the cores for these. Delta-Scuti stars: rotate very rapidly, so modes pushed around a lot. Planets found around them (Murphy 2014, 20165ab, 2018) because their modes are seen to move back and forth - like the pulsar timing detection method. Find a period-luminosity relation for delta-scuti (with the oscillation period relating to the fundamental (n=1) mode; NOTE: these are NOT solar-like oscillations, because at lower frequency and don’t have nice envelope). TESS’s advantage for delta-scuti is large number of these stars scheduled for short-cadence observations (not the case for Kepler). For these observations, can assign n and l to every peak in the power spectrum for these stars - can finally due delta-Scuti asteroseismology! Gaidos comment: Beta Pic is a delta Scuti!! Would be great to get an asteroseismic mass for it. Also combining with some spectroscopic data for these delta-Scutis.

Bill Chaplin - summary and future challenges of asteroseismology and exoplanets. Future challenge - how low can we go down the main sequence? What’s a low mass for an asteroseismologist? < 1 M_Sun, but we do have detections down to K dwarfs. Aspiration: get down to M-dwarfs. Asteroseismic yield from Kepler: 700 solar-type stars (limited by short-cadence slots; includes > 100 planet-hosting stars) and 20000 red giants. On the cool end (< 2 L_sun). Need to detect oscillations with background that includes, from short to long frequency: shot noise, granulation, instrumental artefacts. K-dwarf example: Kepler-444 (highest density dwarf with detected amplitudes). The issue now with TESS, since we have the 2-minute cadence for more stars: low-mass main sequence stars have low oscillation amplitudes. Note: the sampling rate of the observations really matters (especially at high frequency - Nyquist sampling issues), as does having a long baseline. Kepler didn’t detect any oscillations of M-dwarfs because of the limited short-cadence slots, NOT because of the intrinsically low oscillation amplitude. Looking ahead to PLATO. Also, if large enough telescope, can observe power spectrum in Doppler velocity - will get higher SNR than in photometry. What about brown dwarfs? Look for binaries with a brighter primary; assume same age to test if it fits brown dwarf evolutionary models. Done this with one brown dwarf system. The summary - asteroseismology has direct input into exopanets via fundamental properties (density, logg) and dynamical parameters (inclination, rotation, etc.) and indirect input via improvements in stellar physics and modeling (overshooting, mass loss, etc.) and the analysis techniques (i.e. use of model grids).

Allison Youngblood - Ultraviolet Emission from M-dwarfs during quiescence and flares. Stellar activity is most pronounced in UV/X-ray, because emission almost entirely from magnetically heated regions. Even if quiet in optical, can be active in UV/X-ray. Photospheric models of M-dwarfs under-predict flux in UV. Lyman-alpha is important but hard to observe because Earth emits a lot of it (so from ground is hard, but come progress being made) and the ISM severely attenuates it (even to Proxima Cen, core is 100% absorbed) … and need competitive Hubble time (only instrument to observe in that wavelength). M-dwarf UV surveys: Mega-MUSCLES (“inactive” Ms), FUMES (young M’s). Key: all M-dwarfs appear to be UV-active!! Missing EUV (100-900 Angstroms) because of ISM absorption, so need very closest stars. So proposed SMEX 2019 mission ESCAPE to get 50-500 Angstroms for closest stars (sample of ~ 200 distributed over F-M and ages). Flares: how does stellar spectrum change during flares? What are flare frequencies and energies? What kind of data products are useful for exoplanets. Loyd 2018 - M-dwarf FUV emission may be dominated by flares. Is same true for NUV, EUV? Spectra: FUV flares show strong line enhancement and hot blackbody emission; NUV flares are rare. CMEs? Energetic particles have a greater effect on atmospheric chemistry than flare photons (Tilley 2019) … but haven’t detected CMEs for stars yet!!!! Why not? Flare blueshifts are ambigous; Type II radio bursts not detected; but maybe can observe coronal dimming? What is it? Emission lines that trace coronal emission will decrease during CME: CME mass correlated with dimming depth and CME speed is correlated with dimming slope.

Ekaterina Ilin - Stellar magnetic evolution: flaring activity in K2 Open Clusters. We know that flaring, age, and mass are related: Hilton 2011, Chang 2015, Clarke 2018, Howard 2019. She’s adding data to the flaring-age-mass relation using K2 open clusters (M35, M67, Rup147, Praesepe, Hyades, Pleiades). Flare detection using ALTAIPONY, her (open-source) software. Found cumulative flare frequency distributions binned by effective temperature per cluster. Find: 1) flaring rate decreases as Teff increases; 2) all flaring rates decrease from 100 to 600 Myr; 3) rates decrease more for hotter stars.

Eric Gaidos - The star-protoplanet connection in the era of Gaia and TESS. His summary of Kepler results: many small, close-in planets; M-dwarfs have many more of these planets; small planets are gas poor or gas rich; some might be habitable. Gaidos wants age from Gaia, and X-ray luminosity over time (scatter in X-ray luminosity vs. Rossby number - is it due to inter- or intra-star variability?). So look at young clusters within 300 pc: Upper Sco, Lupus, Taurus, NCG 2451, IC 2391, IC 2502, Hyades, Praesepe all observed by K2. David+ 2019: planets in these young clusters are larger. Young M-dwarfs: their pre-main sequence contraction time >> disk lifetime so no disk locking; convective turnover time is long; no pre-main sequence Henyey track, so luminosity monotically decreasing. He’s “collecting” these young M-dwarfs: Gaia + TESS = Age (for pre-main-sequence through isochrone estimates) + Rotation period. Also look at young moving groups - dozens of these. Gaia is adding new members to these groups and adding new moving groups (Faherty 2019, Tang 2019)!! TESS is detecting planets around young moving group stars. Also finding “dipper” stars < 10 Myr - dips can be periodic (typically period ~ a few days … likely inside the dust sublimation radius!!) or episodic (i.e. not predictable) with depths > 10% and variable non-transit shapes. Likely are small masses of dust. Example: HD 240779 (is also a binary - created TESS difference image and found centroid over time). With galactic space motion and lithium abundance, find a ~125 Myr age for it. Where is this dipper-causing dust coming from? The disk, maybe from edge of inner disk? However, many dipper stars have non-inclined disks!! So warp? Also some have no inner disks at all!! So what is it?!?! Younger analog to Tabby’s star … disintegrating exocomets? Dusty winds? Stay tuned - is still a mystery. Note: Beta Pic is also a dipper star (an example of which was found by TESS).

Tuesday, May 21

Elisabeth Newton - Evolution of rotation and activity in M-dwarfs. Fast-rotating M-dwarfs are still fastly rotating at 500 Myr (age of the Hyades), so measuring field M-dwarfs is useful. Find a gap around 30-day periods in the field star stellar mass - rotation period diagram for late-type M-dwarfs. Is it real? Also see gap in activity distribution - different biases, so likely real. Also when overplot clusters give snapshots of these at specific ages (Pleiades, Praesepe) - see a transition from the fast to the slow sequence as a function of age, so it just takes the late Ms a long time to spin down. How long? Based on Galactic dynamics, can divide field dwarfs into young-ish and old-ish groups; see that the transition occurs around 2 Gyr. H-alpha vs. Rossby number shows activity-rotation relation. There’s saturation regime, then activity decreases at higher Rossby number, with scatter of ~1.8x in the relation. At a given Rossby number, more active stars also tend-ish to be more photometrically variable. Stellar rotation and habitable planets: rotation periods are about the same as planet orbital periods for M2-4’s, so will be more difficult to find habitable planets around those stars … this is what happened with Gl 581.

Raphaelle Haywood - what can we learn from the Sun? What’s the limiting factor in our mass measurements? Best precisions are 15-30%. On the Sun, use spatially resolved images from the Helioseismic & Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO), compare to HARPS-N solar telescope, which observes the Sun as a point source (it scrambles all of the spatial information). With >26,000 observations, see photon noise scatter 40-50 cm/s … and RV variations of 8 m/s peak-to-peak!! This is with no planets - significant activity signal! Rotation signal in RVs is quasi-periodic - really confuses detection of 5 M_Earth planet in 25-day orbit. When bin this signal once/day and compare with SDO images, see that they are very compatible. The main deviations between the two are due to granulation/supergranulation that operate at different timescales between the two observations. So their model works, and they see that suppression of convective blueshift is dominant mechanism that produces observed activity-induced RV variation … so, old, slowly rotation stars like the Sun are faculae-dominated (where these plagues are). Is there a reliable activity indicator for these? Based on SDO images, see that there’s a really good correlation between the amplitude of the magnetic field and the observed RV variations … when account for this, reduce the RV variations of the Sun by about half to 55 cm/s. Residual noise is during times when the Sun is most active … so need to measure unpolarized magnetic flux in Zeeman broadening too.

Cecilia Garraffo - stellar spin to planetary atmospheres. Start out talking about gyrochronology; Skumanich relation of rotation period ~ time^¹⁄₂ depends on assumption of dipole … but based on Zeeman doppler measurements of other stars, they’re typically not dipoles. So what does more complex magnetic field structures do to angular momentum loss? From dipole to quadrupole, lose 1 order of magnitude of angular momentum loss efficiency … and 3 OoM of go to even more complex structures. So the more complex the structure (like fully convective M-dwarfs?), the longer it takes for the star to spin down. She developed a complete spin-down model that can reproduce the two branches that Elisabeth talked about, arising when incorporate the complexity of some stellar magnetic fields. Model involves magnetic moments that are an exponential function of Rossby number (rotation period/convective turnover time). Again activity-Rossby number relation: saturation at low Ro (higher convective turnover time) where star reaches maximum activity level. Later-type stars have larger convective regions -> longer turnover time -> lower Ro -> higher activity. So how does this affect habitability of planets around these stars? TRAPPIST-1 lies in the saturated regime; Proxima Cen has Ro ~ 1 which is in increasing regime; Barnard has Ro ~ 3, which is at very low activity. Model the solar wind from these stars and its strength at the planets’ orbits; get stellar wind pressure as a function of planetary orbital phase. For Proxima, pressures get up to 10^3; for TRAPPIST, gets up to 10^5!! If TRAPPIST planets are magnetic, the stellar winds are so strong that the magnetic field lines from the planet connect directly to the star … NO protection happening there, everything from the star gets funneled directly to the planet!!!!! “It’s brutal!” heh … Take-home point: planet-star distance is more important for habitability than activity of the star, because density of the wind is most important, and scales more strongly with distance than with activity. She’s also involved in SMEX concept ESCAPE. From audience questions: note that Zeeman Doppler Imaging resolution depends on the rotation period of the star … important observational bias to account for.

Jennifer van Saders - understanding the limits and potential of period-age-activity relations. Showing a band (as a function of mass and age) from 1.3 - 0.4 M_Sun and > 1 Gyr where period-age relations are ok to use (lower bound increases to 3 Gyr for 0.4 M_Sun). At 1 M_Sun, upper bound of band is 5 Gyr - above there’s inhibited magnetic breaking that makes gyrochronology unreliable. Whether you use empirical or semi-empirical (drawing functional forms from theory), you rely heavily on calibrators for your gyrchonology relation. Problem 1: few calibrators (clusters) above 1 Gyr (being addressed by focused observational efforts). Problem 2: most of these clusters are solar metallicity; it could be an important factor in these relations; will be addressed by Gaia moving groups that span larger range of [Fe/H]. Problem 3: very slowly rotation M-dwarfs aren’t reproduced by models; K-dwarfs might be a problem too (upper envelope for 600-1000 Myr clusters overlap for the K-dwarfs, but would expect there to continue to be spin-down … maybe contamination by field stars? Core-envelope coupling?). Prot - mass relation for field dwarfs: upper edge of relation isn’t as high or sharp as it should be: observational bias, or the inhibited magnetic breaking? Another gap (in addition to and different from the M-dwarfs one) exists for much higher masses (this gap is at longer Prot than the Mdwarf one). Are they different populations? Misidentified rotation periods (¹⁄₂ true Prot)? A critical observability region where spots and faculae cancel either other out? Or don’t have right physics in our model? Take-home: be aware of the region in which gyrochronology is valid! Lots of discussion about the inhibited magnetic breaking regime.

Ruth Angus - Measuring stellar ages from light curves. Start with color-magnitude diagram produced by Gaia … can’t just use isochrones to get ages for planet hosts (which are mostly on main sequence where tracks heavily overlap). Photometric lightcurves are very information-rich: contains rotation, magnetic activity … how to summarize these light curves to pull out these values? (Or map lightcurve -> activity/rotation directly using deep learning?) She’s using autocorrelation function to get Prot, combined with Gaia … clearly see binary sequence, which is confounding factor. Following on Jen’s talk, use gychronology + activity for middle-aged FGK; for subgiants & hot stars, use isochrones; for M-dwarfs, gyrochronology+activity, but those relations are not yet well calibrated. Optimally you’d have one integrated relation that incorporates the right information in the right region in the color-magnitude diagram. How do this? More information in isochrones likelihood where it’s good; more information in gyrochronology where isochrones bad, so actually just combine likelihoods and it naturally balances the two. Code is called “stardate” - selects best dating method in different regions based on the existing information, includes rotation period as a function of all stellar parameters, built on top of Tim Morton’s isochrones package. Calibrated to Praesepe (wide range of stellar masses, data from Douglas 2017) and the Sun. Where gyrochronology doesn’t apply, you’ll get a high variance in the output (so you won’t get a wrong answer, you’ll just get an uninformative one). Also checked with two NGC clusters at 1.1 Gyr, 2.5 Gyr; for 2.5 Gyr cluster, really improve age estimates of G dwarfs when incorporate gyrochronology. Also incorporate stellar kinematics from Gaia. For M-dwarfs, also using activity as an age proxy (work in progress).

Panel discussion - Interesting idea posed by Gaidos: if we can predict where in the activity cycle a star is, then should we wait until the star is at a minimum before measuring planet masses? Elisabeth & Ben: there is a relation between rotation period and activity cycles, issue is what phase the star’s at in its cycle.

Smadar Naoz - friends of planetary systems. The TESS connection: the brightest stars include A stars and subgiants (some are hold, some are cold, some are young, some are old). Majority of these are in binaries. Planet’s orbit evolves as star evolves (main sequence -> giant, tides become more efficient, then lose stellar envelope, the planet orbit expands). Stephan+ 2018: A stars are destroyer of worlds (eccentric Kozai-Lidov + tides very efficient at having planets plunging into stars). Application to a new TESS planet and to KELT-9b: right before engulfment, a farther-out Jupiter-sized planet becomes a “temporary hot jupiter” that these planets may be. What if there’s a second planet orbiting the primary? If they exchange angular momentum faster than the outer stellar companion? Is stable. What if timescales are comparable? Seem stable until orbit crossing. What if outer companion transfers L very fast? Immediately unstable. Dividing line between difference regimes are track in outer planet’s semimajor axis vs. eccentricity of stellar companion. Applying this to Kepler-56. Also suggesting that the low branch in the Prot vs. stellar mass diagram are stars that have engulfed their planet - clear prediction: they should have a companion!!! Claiming that 40% of stars in the fast rotating branch have companions and jupiters and could be explained via this mechanism.

Dong Lai - forming short-period planets: low-e/high-e migration + tidal dissipation. Both hot jupiters (P<10 days) and ultra-short-period planets (P<1 day). High-e migration: advantages is that it can explain hot jupiter pile-up at a few Roche radii, explains why hot jupiters tend to be lonely; also explains high obliquity between stellar spin and planet orbit. Challenge: what mechanism is causing the dissipation - with tides, must assume that planet is 10 times more dissipative than Jupiter; hard to produce P>5 days; also doesn’t produce enough hot jupiters. Dynamic (chaotic) tides significantly resolves these issues. How does it work? Near pericenter, tidal potential of star excites oscillation modes of the planet. Energy transfer during pericenter depends on the oscillation phase of the mode, so need to evolve complex mode amplitude and orbit simultaneously. Different ways the mode energy evolves: quasi-periodic and quasi-diffusive (chaotic mode growth until reach critical threshold, when nonlinear dissipation kicks in). The transition between the two happens at small pericenter distance and large e. When introduce dynamical tides, the decay starts immediately at first pericenter passage - much faster than traditional Kozai cycles + regular tidal dissipation. It also saves some planets from tidal disruption because the strong dissipation truncates how high the eccentricity gets (producing more hot jupiters in the end). Also more easily produce hot jupiters at 5 days. Secular Orbital Chaos (between 3,4,5 giant planets) also produces high-e migration of innermost planet when there’s a sufficient angular momentum deficit and when secular resonances exist and overlap; if you don’t have dynamical tides, all of these high-e planets will be tidally destroyed. Except secular chaos doesn’t produce > 5 day planets easily because the eccentricity evolution is too fast. Next steps: what happens to the planet with tidal heating? Ulta-short-period planets: check out Winn 2019 for review on these. Are they a separate population? maybe - period distribution is different, have different size distribution (no radius gap), systems with USPs have larger mutual inclinations (have fewer co-transiting companions). He suggests forming them via low-e migration + tidal dissipation in the planet. What is low-e migration? Start with a multiple planet system (requirements; outer planets (at least 2) slightly more massive than inner one (2-3x) that are slightly eccentric (0.05-0.1), inner planet must start at 1-3 days) where they “share” eccentricity (must be > a few % to get Gyr timescales), tidal dissipation happens gradually. This excites mutual inclinations, and the “final distribution of USPs produced agrees with observations under wide conditions” - didn’t have time to show evidence behind this claim.

Konstantin Batygin - hyperbolic encounters. Start out talking about stars forming in clusters, but hard to tell if formed in small or large cluster (some population arguments). Also talking about the Kuiper belt (the “Duncan donut” lolz), within which is the cold classical belt (free inclination dispersion < 1.7 degrees when correct for observation angle) … how to keep this belt so cold? Starting from scratch, developing secular theory for cluster interactions. Have two components: mean cluster potential (~ static) and stellar flybys (stochastic). The potential part: orbit-averaged Hamiltonian is reminiscent of the Kozai-Lidov Hamiltonian, so it’s a generalized Kozai effect. The more complicated part is the stellar flybys. He’s discarding the usual approach of impulse approximation because stars are moving too fast compared to the cluster crossing time … so encounters are secular, and the gravitational perturbations happen to a ring … and the quadrupole-level expansion of the Hamiltonian overaged over the particle’s orbital period is ugly. Approximation simplifies things and reduces this to harmonics: effective secular evolution is slow compared to crossing time. Terms depending on e go to zero, and get the hamiltonian of a pendulum - it’s hyperbolic. Applied this to Orion nebula-like cluster, compared to N-body results. In a statistical sense they are very comparable. The inclination dispersion increases with time is almost linear in log space, steeper than slope = ¹⁄₂ so not a diffusive process.