The effects of stellar activity on detecting and characterising exoplanets Suzanne Aigrain R. Angus, J. Barstow, V. Rajpaul, E. Gillen, H. Parviainen, B. Pope, S. Roberts, A. McQuillan, N. Gibson, T. Mazeh, F. Pont, S. Zucker Timescales Solar irradiance (SoHO/VIRGO) Transits are easy to separate from Active regions photometric variations due to star spots … up to a point! Activity-induced variability is more problematic for: - transmission spectroscopy - radial velocity planet searches - phase curve studies Granulation Orbit Transits Oscillations (Aigrain, Favata & Gilmore 2004) Filtering activity to detect transits Iterative non-linear filter followed by least-squares box-shaped transit search (Aigrain & Irwin 2004) P ~ 20h, depth 0.0003, Rplanet ~ 2 REarth (CoRoT-7b, Leger et al. 2009) When does activity matter for transit searches? Transit SNR = sqrt(Ntransits) x depth / sigma(Ttransit) where: Intrinsic stellar variability on 6 hour time- - Ntransits is number of transits scales from Kepler (Gilliland et al. 2011) - Ttransit is duration of transit Activity means Kepler would have needed 7 rather than 4 years to reach SNR of 10 for Earth-like planets in the Sun habitable zone of Sun-like stars This can be addressed, at least partially, by modelling the activity-induced variations simultaneously with the transits Modelling stellar signals and instrumental systematics jointly using Gaussian Processes (GPs) Model activity as a quasi-periodic a Gaussian process. Example from K2 Campaign 7 Simultaneously model pointing-related systematics K2SC pipeline - Aigrain et al. 2016. Code and LCs available - talk to Hannu Parviainen First planet candidate catalog: Pope et al. (2016, on arXiv this week) Activity in transit spectra • In any kind of high precision transit studies, need to worry about spots (see e.g. Pont et al. 2013): • occulted: distort transit, or make it seem shallower • un-occulted: make transit appear deeper • Both effects are very important and hard to correct for transmission spectroscopy • even in the IR (see e.g. Barstow et al. 2015 - JWST) • Plages may also be important (low contrast but large area) - Oshagh et al. (2014) Spectroscopic effects of star spots Contrast between 5000 K photosphere and cool spots with different temperatures (MARCS models, Gustafsson et al. 2008, log g = 4.5, [Fe/H] = 0) TiO / VO Tspot = 3500 K H2O Mg Tspot = 4750 K Na Accounting for spots in transmission spectra 4 D. K. Sing et al. Estimatecoefficients with spectrum/temperature 1D stellar atmospheric models andof finallyspots using from a fully 3D time-dependent hydrodynamic stellar atmospheric model. 8 D. K. Sing et al. occultedWe computed spots limb-darkening coefficients for the linear law HD189733b, I(µ) 1 u(1 µ)(1) Sing et al. (2012) I(1) = − − as well as the Claret (2000) four-parameter limb-darkening law Tspot ~ 4250K - unique way to I(µ) 1 c (1 µ1/2) c (1measureµ) spot temperatures on I(1) = − 1 − − 2 − c (1 µ3/2) c (1 µ2). (2)other stars! − 3 − − 4 − 8 D. K. Sing et al. For the 1D models, we followed the procedures of Sing (2010) Figure 3. STIS white light curve for visit 1 (black) and visit 2 (red) with using 1D Kurucz ATLAS models2 and the transmission function the instrument trends removed. The points showing occulted spot features of the G430L grating (see Table 1). The four-parameter law is the are indicated with boxes. The best-fitting transit models for both visits are best representation of the stellar model intensity distribution itself, shown in grey using the unspotted points. while the linear law is the most useful in this study when fitting for 3.1 Instrument systematic trends the coefficients from the transit light curves. We constructed a 3D time-dependent hydrodynamical model at- As in past STIS studies, we applied orbit-to-orbit flux corrections mosphere using the STAGGERCODE (Nordlund & Galsgaard 1995) with by fitting for a fourth-order polynomial to the photometric time aresolutionof2403 grid points, spanning 4 4Mm2 on the hori- series, phased on the HST orbital period. The systematic trends zontal axes and 2.2 Mm on the vertical axis.× The simulation has a were fit simultaneously with the transit parameters in the fit. Higher time-average effective temperature Teff 5050 K, surface gravity order polynomial fits were not statistically justified, based upon log g 4.53 and metallicity [Fe/H]⟨ 0.0⟩ = (based on the solar com- the Bayesian information criteria (BIC; Schwarz 1978). Compared position= of Asplund, Grevesse & Sauval= 2005), which is close to the to the standard χ 2, the BIC penalizes models with larger num- stellar parameters of Bouchy et al. (2005). Full 3D radiative transfer bers of free parameters, giving a useful criterion to help select was computed in local thermodynamic equilibrium (LTE) based on between different models with different numbers of free param- continuous and spectral line opacities provided by Trampedach (in eters, and helps ensure that the preferred model does not over- preparation) and Plez (private communication); see also Gustafsson Figure 8. Plotted is the occulted-spotfit the data. The feature baseline from flux level visit of 2 each at (top visit to was bottom) let free to et al. (2008). We obtain monochromatic surface intensities Iλ(µ, φ, 3300, 3950, 4450, 4950 andvary 5450 in time Å along linearly, with described the best-fitting by two fit parameters. spot solution. In addi- x, y, t)withasamplingofλ/%λ 20 000 in wavelength for 17 tion, we found it useful to also fit for further systematic trends polar angles µ,fourazimuthalangles= φ and a time series of 10 which correlated with the detector position of the spectra, as de- snapshots that span 30 min of stellar time t, with a horizontal res- Figure 10. Spectral≈ signature of the stellar spots occultation derived from Figuretermined 8. Plotted from a is linear the occulted-spot spectral trace feature in IRAF fromand visit the 2 dispersion- at (top to bottom)olution of 120 120 grid points in x and y. Limb-darkening laws × 3300,direction 3950, subpixel 4450, 4950 shift and between 5450 Å spectral along with exposures, the best-fitting measured spot by solution.theI(µ)/ STISI(1) (see G430L Fig. 4) (closed were derived black by and averaging green symbols)Iλ(µ, φ, x, y and, t) ACS (open symbols) cross-correlation. data.over horizontal The spot grid, is azimuth modelled angle and with time, stellar providing atmospheric a statistical models of different We found that fitting the systematic trends with a fourth-order representation of the surface granulation, and integrating over each temperaturesFigure ranging 10. Spectral from signature 4750 toof 3500the stellar K in spots 250 occultation K intervals derived (blue from to orange, polynomial HST orbital period correction and linear baseline limited bandpassthe and STIS the G430Ltransmission (closed function; black and the green result symbols) was normalized and ACS (open symbols) S/N values to the range of 9000 to 10 000 (precisions levels of 0.011 respectively),to the disc-centredata. The and intensity spotTeff is at modelledµ 50001. A withK more for stellar detailed the stellar atmospheric description temperature. modelsof of different = = to 0.01 per cent). These limiting values match similar previous the 3D modeltemperatures will be given ranging in from a forthcoming 4750 to 3500 paper K inby 250 WH. K intervals (blue to orange, pre-SM4 STIS observations of HD 209458b, which also similarly Whilerespectively), limb darkening and isTeff stronger5000 at K near-UV for the stellar and blue temperature. wave- corrected for these systematic trends (Brown 2001; Ballester et al. lengths,As in compared Pont et to the al. red (2008), and= near-infrared, we model the white the effects light stel- of unocculted spots 2007; Knutson et al. 2007b; Sing et al. 2008a). With the additional bylar intensity assuming profile that is also thepredicted emission to be close spectrum to linear (see Fig. of 4). spots corresponds to correction of position-related trends (see Fig. 1), we were able to Fortunately,As a linear in Pont stellar et al. intensity (2008), profile we model makes it the much effects easier of to unocculted spots increase the extracted S/N to values of 14 000 per image, which acompare stellarby fit limb-darkening spectrumassuming that of coefficients the lower emission to temperature model spectrum values, because of than spots it the corresponds rest of to the star is 80 per cent of the Poisson-limited value. These additional free is less ambiguousa stellar asspectrum to which of limb-darkening lower temperature law to choose than the and rest of the star ∼ covering a fraction of the stellar surface and assuming no change in parameters in the fit are also justified by the BIC as well as a reduced becausecovering the fit is not a fraction complicated of the by degeneracies stellar surface when and fitting assuming for no change in χ 2 value. In a fit excluding the position-dependent systematic trends the surface brightness outside spots. We neglect the effect of faculae ν multiplethe limb-darkening surface brightness coefficients. outside In the spots. white We light-curve neglect the fits, effect of faculae for the first STIS visit, we find a BIC value of 277 from a fit with 73 onwe performed the transmission fits allowing the spectrum. linear coefficient The term spots to vary freely,then lead to an overall 2 on the transmission spectrum. The spots then lead to an overall degrees of freedom (DOF), eight free parameters and a reduced χ ν as well as fits setting the coefficients to their 1D and 3D predicted of 3.31. Including the position-related trends lowers the BIC value dimmingvalues withdimming the of four-parameter the of star.the star.