Radial Velocity Planet Surveys. Jian Ge, University of Florida

Transcription

1 Radial Velocity Planet Surveys Jian Ge, University of Florida 1

2 Theory vs. Observation Ida & Lin,

3 Major efforts for detecting new planets since 1989 ( Doppler method (386 planets) Photometry method (64 planets) Direct imaging (11 planets) Microlensing method (10 planets) Radio pulsars (9 planets) 3

4 Latham et al Wolszczan & Frail 1992 Mayor & Quolez, 1995 Exoplanets discovery history ( Breakthrough in Doppler, Transit, Imaging & Microlensing 4

5 Doppler RV Method The first planet discovered in the solar neighborhood in

6 The discovery of a 0.5 Jupiter mass planet around 51 Peg in 1995 by Mayor & Quolez has triggered exoplanet revolution Radial velocity data 1.93 meter telescope in Haute-Provence Observatories in France 6

7 Doppler Wobble Physics Observer V + M p M * V 28.4 P 1/3 M P M sin 2/3 * i m/s a p a s M P in Jupiter masses, P in years and m * in solar masses 7

8 Detectability of planets around solar type stars with Doppler techniques 2 m/s 0.1 m/s From Artie Hatzes Late F, G and K dwarfs (~ solar masses) 8

9 Detectability of planets around intermediate mass stars with Doppler techniques 50 m/s 2 m/s A type MS star, K and G subgiants and giants (~1.2-5 solar masses) 9

10 Detectability of planets around low mass stars with Doppler techniques 10 m/s 1 m/s M dwarfs ( solar masses) 10

11 Detectability of RV planets: Doppler precision ( ) and RV signal semi-amplitude (K) Largely depends on K/, cadence and number of observations if photon noise error is dominant K/ ~2 False Alarm Probability About 50% planets with K/ ~2 and relatively low orbit eccentricity can be detected with ~30 randomly distributed observations over more than 1 period 11

12 Simulation of detectability of a 2 Earth mass 2 day planet around a solar type star with 1 m/s Doppler precision K/ ~1 False Alarm Probability K= 1m/s, = 1m/s K/ ~1 requires ~ measurements to have better than 50% detection efficiency 12

13 Doppler Techniques High resolution cross-dispersed echelle spectrographs (~1950 s present, proposed by Struve 1952) Dispersed fixed-delay interferometer (1997 present, proposed by Erskine 1997) 13

14 Cross-dispersed echelle spectrometer Doppler technique work principle Lick Hamilton Cross-dispersed Echelle Spectrograph Entrance Slit Schmidt Mirror Prism Cross-disperser Detector Schmidt Corrector Echelle Collimator From Lick website 14

15 Spectral Format for a cross-dispersed echelle Spectrometer Cross-dispersed spectral format Cross-disperser 15

16 Solar spectrum in m with a crossdispersed echelle Spectrometer From NOAO website 16

17 Doppler quality factor, Q, carried out by different spectral type stars From Bouchy et al

18 Echelle Working Principle for Doppler RV Measurements A normalized intrinsic absorption line: Intensity I =1 I D Slope: I di / d di / d const. D c I Velocity, RV Measurement uncertainty for a line:, i I di / d c D i I F i 18

19 Total Doppler error budget caused by photon noise i 1, N 1 i 2 1/ 2 1/ ) F R e, RV 3/ 2 ( e, i D Total photon number collected by lines F A I t Narrow line and deep line have better Doppler sensitivity Doppler sensitivity is proportional to S/N=F 1/2 Doppler sensitivity is proportional to spectral resolution of R 3/2 19

20 Dispersed Fixed-delay Interferometer Doppler technique Work Principle Telescope Interferometer assembly Fringes Cylinder Fibers Spectrograph with R~10,000 Dispersed Fringes Detector Erskine & Ge (2000), Ge et al. (2002), Ge (2002) Doppler shift: V (phase shift) 20

21 Fixed-delay Interferometer Doppler Principle Incoming beam Fringe order and delay relationship: m d Mirror 2 image Mirror 1 d/2 /2 m m-1 Fringe shift caused by wavelength change: d m 2 c m d d c Doppler shift and phase shift relationship: c d 2 21

22 Doppler sensitivity with fringe modulation I sin(2 ), c / d 1 0 Fringe average slope: 4d c Visibility: I I max max I I min min At photon noise limit: f, i 4d c F i 22

23 Total Doppler error budget caused by photon noise i 1, N 1 i 2 1/ 2 1/ 1/ ) F R f, RV 2 ( f, i D Narrow line and deep line have better Doppler sensitivity Doppler sensitivity is proportional to S/N=F 1/2 Doppler sensitivity is proportional to spectral resolution of R 1/2 23

24 Echelle method vs. DFDI method Doppler sensitivity comparison: fringe, ob echelle, ob fringe 1/ 2 3/ 2 echelle : spectral resolution Echelle requires high spectral resolution, such as R~60,000 to reach high Doppler sensitivity DFDI method can reach high precision with moderate spectral resolution, such as R ~ 10,000 Multiple object capability and high throughput can be realized by DFDI method 24

25 High resolution echelle vs. dispersed FDI Stellar echelle spectrum x comb FDI spectrum Credits: Julian van Eyken 25

26 Instrument Drift Calibration Method I: Absorption cell Superimposing of rest from absorption lines on stellar lines: HF and iodine in the optical The most popular one: iodine absorption lines m 26

27 Modeling the Observations The observation is modeled as the product of two functions, the intrinsic stellar spectrum, I S, and the transmission function of the iodine absorption cell, T 12 and convolved with the spectrograph PSF and binned to the wavelength extent of the CCD pixels. I obs ( ) k T ( ) I ( ) * PSF 12 s where k is a normalization factor, is the Doppler shift, and * represents convolution. 27

28 Process for determining Doppler shift in stellar spectra using the echelle and iodine technique (Butler et al. 1996) Iodine template stellar template Synthetic spectrum Residual The Doppler shift is determined by comparing the model to the observation using 28 a standard Marquardt non-linear least squares algorithm

29 Best Doppler precision with echelle plus iodine at Keck with HIRES with R~80,000 (Vogt et al.) Long term Doppler precision is around 1 m/s 29

30 Instrument Drift Calibration Method II: ThAr separation beam calibration Thorium lines 30 From Mayor 2005

31 Operating HARPS on the ESO 3.6m telescope in temperature controlled vacuum chamber Instrument long term stability: T =0.01 K P=0.01 mbar From Mayor

32 HARPS Thermal stability Stability during one day: K rms Stability during one year: <0.01 K 32 From Rupprecht et al., 2004

33 RV tracking performance with simultaneous ThAr reference 33 Mayor et al. 2003, The ESO Messenger

34 RV Observations of HD with HARPS Three planets around HD (a K0V, V=5.75) Overall rms for the RV residuals is 0.81 m/s: earlier measurements with 1.5m/s and later with 0.64 m/s (Lovis et al 2006) 34

35 Advantage and disadvantage of these two calibration methods Iodine: Robust and nearly independent of environment changes Long term stability Reliable statistics on planet properties Limited wavelength coverage, only m ThAr: Large wavelength coverage potential much better precision and suitable for different spectral stars Easy data processing Required very high thermal and pressure stability for high precision and long term stability Less reliable for long term calibration and possible more long term systematic errors 35

36 Planet mass distribution from California Group Less certain at less than 0.2 Jupiter mass planets due to Doppler sensitivity Less certain in brown dwarf desert due to relatively small survey sample / / / / / / / / / Marcy et al

37 Extrasolar planet orbital distribution from California Group 3 day pile-up incomplete Marcy et al There appear to be a dip in the orbital distribution between a few days to 1 year 37

38 Exoplanet orbital eccentricity distribution from California Group Marcy et al Most of the extrasolar planets are in the eccentric orbits, unlike the solar system planets! 38

39 Eccentricity distribution fall off at high value Hatzes et al Fall off at high value due to observation bias? Or real? 39

40 Multiple planet systems from California Group 28 well defined multiple planet systems (Wright et al. 2009) 4 of them are in the motion resonant About 28% planet systems have multiple planets including long term trends Planets in multiple systems have less eccentric orbits Single planet system shows 3 day pileup while multiple systems show a more homogenous distribution in log-period 40

41 Planet Metallicity Correlation (Fischer & Valenti 2005) 2 P planet ~ ( N Fe / N H ) Abundance analysis of 1000 stars on planet search. Flatten out? Fischer & Valenti 2005 Low metallicity giant planets may form through a different mechanism from that for 41 high metallicity ones

42 Planet occurrence versus stellar mass for giant planet within 2.5AU (Johnson et al. 2007) Planet occurrence strongly correlated with stellar masses 42

43 Major Limitations for RV Planet Surveys Stellar intrinsic limitation Photon noise limited to relatively bright stars and solar type stars IR Doppler technique for low mass stars Stellar activity (about a few m/s to ~100 m/s) Photometry, Bisector variation and Ca II emission IR observations Binary stars Multiple dimensional cross-correlations (Mazeh et al.) Acoustic modes (asteroseismology) Integrate over 15 min to average out RV variation 43

44 Major Limitations for RV Planet Surveys External limitation Small sample due to single object capability, only about ~3000 stars well observed over the last 14 years at a dozen telescopes Multiple object surveys Wavelength calibration issue with iodine and Thar Astro laser combs cm/s precision Telescope guiding Fiber mode scrambling 0.1 m/s possible Stability in the illumination of spectrograph Instrument environment control, pupil mask etc Detector-related effects Careful modeling, sampling a resolution elementt by more than 2 pixels 44

45 Next Generation Extrasolar Planet Surveys Multiple object exoplanet surveys to largely increase survey sample (including intermediate mass stars) Example: Multi-object APO Radial Velocity Exoplanet Large-area Survey (MARVELS) Global high precision Doppler instrument network to increase RV sample and better handle stellar noises Example: EXtremely high Precision Extrasolar planet Tracker (EXPERT) network Infrared Doppler planet surveys to include low mass stars and young stars Example: Florida IR Silicon immersion grating spectrometer (FIRST) 45

46 MARVELS Survey To monitor a total of 10,000 V= FGK dwarfs and subgiants, & 1,000 V= G and K giants with minimal metallicity and age biases for detecting and characterizing ~200 giant planets using SDSS telescope in Use all of the bright time in and share the bright time with APOGEE in Each of ~120 fields will be monitored about 30 times over ~18 months Two multi-object Doppler instruments with a total of 120 object capability The wavelength coverage ~ nm Spectral resolution ~10,000 Doppler precision (photon noise limit) in 1 hour exposures: 3.4m/s (V=8), 8.5 m/s (V=10) and 21.3 m/s (V=12) 46

47 The MARVELS Survey Science Goals Principal science goals: find a homogeneous sample of several hundreds of giant planets that can be used for statistical study of planet properties and comparison to theory constrain formation, migration & dynamical evolution of planetary systems discovery of rare systems (e.g. Very Hot Jupiters, short-period super-massive planets, short-period eccentric planets, transiting planets, highly eccentric planets, rapidly interacting multiple planet systems, planets orbiting low-metallicitiy host stars, planets around active and young stars, and other rare types of planets) signposts for lower-mass or more distant planets quantify the emptiness of the brown dwarf desert 47

48 MARVELS Survey Target Selection Effective temperature and gravity plot of target candidates of 7 fields Targets initially selected from GSC2.3 catalog matched up with 2MASS for J, H, K colors Metallicity distribution SDSS spectroscopy preselection efficiently remove giants (e.g., log g < 3.0) and stars that are too hot (e.g., T eff > 6250 K) from the initially color selected targets 25-35% of the ~500 candidates per field are acceptable targets 48

49 Distribution of POST-Selection T eff of MARVELS Targets ~30% of our sample between G2V-F8V, ~70% are later than G2V 49

50 Final Distribution of Targets Final distributions based on 41 of the 60 first season fields ~90% of MARVELS targets having V < 11.5 ~55% main sequence stars, ~35% subgiants and ~10% giants 50

51 The Multi-object Optical Doppler instrument at SDSS in Sept 08 MARVELS Plugging Plate MARVELS-I inside an air condition room Calibration box Instrument MARVELS Fibers Control box 51

52 A full frame of a ThAr spectrum with MARVELS instrument 120 spectra from 60 fibers A total of 120 ThAr spectra occupy middle 4000x4096 pixels of the 4kx4k 52 CCD

53 A full frame of 120 stellar fringing spectra from the HAT-P-1 field in 40 min with MARVELS instrument Dispersion direction 53

54 MARVELS instrument RV drift Zoom in 54

55 Doppler Measurement Precision over 40 days RMS~ 3m/s 55

56 Discoveries of Two Brown Dwarfs by MARVELS A new brown dwarf with 20 Jupiter masses and 5.9 day period A new brown dwarf with 50 Jupiter masses and 5.8day period Lee et al. 2009, in preparation Fleming et al. 2009, in preparation 56

57 Two Example Planet Candidates by MARVELS A short period planet candidate with P=8.8 days, e=0.3, Msini=1.5M J An intermediate period planet candidate with P=25.1 days, e=0.1, Msini=4.1M J 57

58 RV uncertainty as a function of V magnitude ~75% stars with rms errors within 1-2 photon limiting errors, ~25% with much large RMS due to short period binaries, giant stars, low visibility, outliers etc. 58 Further refining of data pipeline to approach photon noise limiting performance.

59 Current Survey Status Number of Survey Observations: 489 Number of Plates Observed: 43 Total Number of stars:

60 Global Extremely High Precision Exoplanet Tracker Network USA EXPERT Spain SET China LiJET m, R=18,000, m/s in 30 min for V< 8 solar type stars with 2 m telescopes 60

61 EXPERT hardware setup at Kitt Peak 2.1m in October 2009 EXPERT inside a 2.1m Coude room EXPERT control chassis 61

62 EXPERT inside a Thermally Controlled Chamber at Kitt Peak in Sept Thermal enclosure Chamber Echelle Interferometer RV input fiber DEM input fiber Prism Collimator/Camera 4kx4k CCD Pressure sensor 62

63 Solar Spectra on the Detector with EXPERT Order 29, 0.70 m Output 2 Order 52, 0.39 m Order 29, 0.70 m Output 1 Order 52, 0.39 m 63

64 Early RV measurement results with EXPERT Observations with Iodine absorption Observations with sky Photon error =1.3m/s with 12 Photon error =1.2m/s with orders combined 12 orders combined Observations with 51 Peg Photon error =4.7m/s with 9 orders combined 64

65 Photon limited RV precision in 30 min exposures with EXPERT Magnitude RV precision V=6 0.4 m/s V=7 0.6 m/s V=8 1.0 m/s Primary targets for EXPERT network 65

66 FIRST IR high resolution spectrograph Optical Layout APO 3.5m Telescope FIRST with R=55,000, um simultaneously with 2kx2k H2RG array Commissioning in Fall

67 Silicon Immersion Grating for Infrared High Precision Doppler Measurements Dispersion n (100) Top view of UF Silicon Immersion grating (111) (111) 50 mm Same size grating, R increases by a factor of 3.4 Same R, the spectrograph size & weight can be significantly reduced 67

68 IR Laser Comb for Radial Velocity Calibration being Built by NIST, Colorado Power per combline (nw) Spectral Lines Measured in the NIST Lab, Resolved 12.5 GHz modes nm output (resolved 12.5 GHz modes) input (unresolved 250 MHz modes) Wavelength (nm)

69 Photon Noise Limited Doppler Precision with FIRST at the APO 3.5m telescope H = 6, we would need an exposure of about 25 minutes to achieve 1 m/s precision. 69

70 Summary and Future Initiatives Doppler planet surveys have produced ~80% known planets and will continue this trend for many years Doppler planet surveys provide best constraints on inner planets around nearby FGKM stars Single object exoplanet surveys multiple object exoplanet surveys to largely increase statistical sample of planets for tudying planet properties, constraining planet formation and evolution Single object high precision Doppler instrument network of high precision Doppler instruments at different sites Near IR Doppler planet surveys will enable planet detection around low mass stars (including habitable Earth mass planets) and young stars 70