High contrast spectroscopy Michelson Summer School 2004 Bruce Woodgate, Goddard Space Flight Center Topics Spectra of known planets Examples of current attempts for high contrast spectroscopy How to do better Integral field spectrographs Why do spectroscopy? • Spectroscopy is harder than pictures, so why bother? (The public won’t get to see it! ) - For detecting an object, if we don’t introduce much detector noise or light losses, we can get all the photons from an object at once, so it can be faster than serial filter imaging. - For distinguishing an object from background, such as a planet from the scattered light of a parent star, or from background stars, spectroscopic features can identify it. Need not wait years for proper motion. - Characterize the object’s atmosphere: molecules and atoms, temperature, density - Radial velocity, orbit kinematics Examples of planets spectra Spectra of planets (from Karkoschka, Icarus, 1994, 111,174) Neptune spectrum (Melillo, 2000) Earth extinction spectrum (from Karkoschka, Icarus, 111,174) Earthshine spectrum (Woolf et al, ApJ 574, 430) Keck spectrum of molecular hydrogen around eclipsed star HST/STIS slit position for brown dwarf spectrum GL 229b GL 229a (star) GL 229b (brown dwarf) Cool object spectra (Schultz et al, 1998, ApJL 482, L181) GL 229b spectrum taken with HST/STIS Examples of current attempts for high contrast spectroscopy STIS “coronagraphic” format Real instrument uses mirrors CCD detector Focal plane wedge mask blocks star with selectable width Hubble space telescope + corrector mirrors Pupil plane mask covers 15% outer ring Correction for spherical aberration only No mid-frequency wavefront correction No masking of secondary and spider diffraction Gas and Dust cloud condensing around AB Auriga - Hubble/STIS coronagraphic image, PSF subtracted Visible X-rays ↓ HH 409C HH 409A ← new HH knots?→ STIS 1998 Coronagraphic Image of HD163296 (Grady et al. 2000) placed next to Chandra 2003 X-Ray Image (20”x20”) of the same field, courtesy of Doug Swartz (USRA/NSSTC), adjusted to the same plate scale. As HH 409A and 409C have moved out of the field, we believe the x-ray image has revealed one if not two new HH knots. Jets and Older Herbig Ae Stars • The first indication that the conventional wisdom that nearZAMS intermediatemass stars don’t have jets was wrong was provided by HD 163296. HD163296 STIS red and UV spectra RED – G750L H alpha S II UV – G140M Si III Lyman alpha red-shifted Lyman alpha blue-shifted Spectroscopic formats onto 2-D detectors Focal Plane Feed to Spectrograph Echelle (two spectral dimensions) Long Slit (one spatial, one spectral dimension) Tunable narrow band imager (two spatial dimensions, serial tune for spectral) 1 ImageSlicer (two spatial, one spectral dimension) 2 3 4 1 2 3 4 Detector STIS G140M Long Slit Observations over two epochs The 1999 image is discussed in Devine et al (2000). Lyman α observations manifest a proper motion asymmetry between the jet and counterjet, while observations in Si III highlight an asymmetry in ionization. Looking at Si III, it is also evident that the jet is decelerating as it moves away from the star. HH 409C HH 409B HH 409A HD163296 H Alpha Fabry-Perot in coronagraphic mode at Apache Point 3.5-m telescope HD163296 (H Alpha – offband) Fabry-Perot in coronagraphic mode at Apache Point 3.5-m telescope HD100546 – HST/STIS visible coronagraphic image, combining 2 roll angles • Mapping the Environment of HD 100546 Long slit spectrum obtained in October 2001 along the system major axis. Visible UV Lyman alpha – G140M • Mapping the Environment of HD 100546 System minor axis observed in June 2002. HD 100546: the Inner 100 AU Minor axis Major axis • Ly a, H2, and reflection nebulosity is enhanced along the system minor axis to the NE of the star. • Consistent with an origin in the envelope and not in the disk. The HD 104237 Bipolar Jet •G140M long slit spectroscopy of HD 104237 reveals a microjet in Lyman a. •The approaching jet is along PA=342°, with the counterjet along PA=162°. •The approaching jet extends 1.05” from the star, while the counterjet can be traced 2.6” from the star. Planet transiting star – HD 209458 HST/STIS photometry (from Brown et al, 2001, ApJ, 552, 699) Sodium in occulting planet’s atmosphere HD 209458b HST/STIS G750M spectra from Charbonneau et al, 2002, ApJ 568, 377 λ=1750A HST/STIS echelle spectrum - format used for long high resolution spectra - one long spectrum is chopped into many short spectral orders λ=1150A Supernova 1987A red spectra Wide slit spectra (2 arcsec), to show 2 spatial dimensions and one spectral. One spatial dimension is convolved with the spectral dimension How to go deeper • Observe all the wavelengths, and all the relevant spatial points simultaneously • Maintain high efficiency and low background Factors controlling signal to noise (1) The signal to noise per resolution element (S/N) in a measurement is (S/N) = Nsig / (Nsig + Nb)1/2 where Nsig = number of counts from source Nb = number of counts from background = Nscatt + Nsky + Ndarks +Nread Nscatt = residual counts from star (after any suppression) Nsky = counts from sky, airglow, our zodi, circumstellar, interstellar, etc Ndarks = detector dark counts = Cdarksnpt where Cdarks = detector dark rate np = number of pixels per resolution element t = total exposure time Nread = equivalent counts from detector read noise = nrnp(nrms)2 For multiple reads, nr = number of reads = t/t0 where where t0 = max exposure time to remove cosmic rays (nrms)2 = equivalent detector read noise counts per pixel per read Factors controlling signal to noise (2) Converting to rates, N = Ct, (S/N) = Csigt / (Csigt + Cscattt + Cskyt + Cdarksnpt + (t/t0)np(nrms)2 )1/2 (S/N) = Csigt1/2 / (Csig + Cscatt + Csky + Cdarksnp + (np/t0)(nrms)2 )1/2 Renaming, (S/N) = Csigt1/2 / Σnoiserate1/2 Total exposure time, t = (S/N)2 Σnoiserate / Csig For an excellent coronagraphic spectrograph (TPF-C) observing a faint planet, using a cold detector, Csig, Cscatt, Cdarks, are small. Then Csky or Cread = (np/t0)(nrms)2 could be the dominant noise source contributing to Σnoiserate. Read noise is dominant with the best regular CCD (with nrms ~ 2 electrons), if (np/t0)(nrms)2 > FskyAeffΩδλ. [Csky = FskyAeffΩδλ, Fsky (zodi)~2.5x108 ph/(cm2s.sr.μ), Aeff=4.3x104cm2, Ω=1.5x10-14 sr, np=9, t0=1000s] For TPF-C parameters, read noise is dominant for δλ < 0.23μ (R~10). So for spectral resolution elements narrower than broad band filters, zero read noise photon counters are needed. See also Lindler and Heap simulations later. Spectroscopic strategies High contrast spectroscopy 1) Reduce scattering source - coronagraphy, nulling interferometry 2) Reduce other backgrounds (detector, sky) - photon counting detector, high angular resolution 3) Observe source and background simultaneously - subtract background from source under same conditions, eg seeing, thermal, pointing, deformable mirror status 4) Select wavelength, resolution, polarization to improve signal to background - eg UV, narrow spectral band for gas line emission, polarization for scattered light, IR for thermal emission 5) Include as many source photons as possible - integral field spectroscopy 6) Observe background surroundings broadly to estimate background at source position - integral field spectroscopy, separate source and background with spectral template 7) Observe reference point source under similar conditions - PSF subtract TPF Spectroscopic Requirements Wavelength range 0.5 – 1.0 microns Resolving Power ~70 (Nyquist sampled if read noise zero) Spatial sampling (i) Nyquist sample 6.0 meter diffraction limit at 0.5 microns (0.018 arcsec ) (ii) Elliptical: Nyquist sample 6.0 x 3.5 m diffraction limit at 0.5 microns (0.018 x 0.031 arcsec ) Spatial coverage Cover coronagraphic dark hole (= 1.8 arcsec square if have 96 x 96 DM actuators) TPF SPECTROSCOPY - TRADE BETWEEN SLIT AND INTEGRAL FIELD SPECTROGRAPH Property Transmission (using prism) Slit (towards star) ~0.8 IFU ~0.8 Roll alignment for point source Needed Not Needed Alignment (slit to star) Difficult Easy Multiple planets? No Yes Disk spectra Less sensitive More sensitive Help find planets? No Yes, if buried in speckles Dark hole edge Traditional Integral Field Techniques Focal Plane Feed to Spectrograph Detector Lenslet Array Fiber Bundle Image Slicer 1 2 3 4 1 2 3 4 Based on a figure from Content, 1998. INTEGRAL FIELD UNIT OPTIONS Lenslet array – concentrates and separates images to allow room to interleave spectra - examples: CFHT/Tiger, Oasis, SAURON, SNIFS, OSIRIS, MEIFU - smallest and lightest - high throughput - large spatial format - limited spectral elements, use for low spectral resolution or small spectral range 2 Mirror image slicer array – rectangle to line reformat - examples: MEIFU, MUSE, JWST/NIRSpec, SNAP, GNIRS, KMOS, SPIFFI - intermediate weight and size - high throughput - small spatial format - large spectral format Fiber/lenslet array – rectangle to line reformat - examples: SMIRFS-IFU, GMOS, SILFID, INTEGRAL - large and heavy - lowest throughput (fiber light losses) - OK for very wide field ground-based MOS OSIRIS James Larkin (PI), Alfred Krabbe (Co-PI), Andreas Quirrenbach(PS), Sean Adkins, Ted Aliado, Paola Amico, Matthew Barczys, George Brims, John Canfield, Thomas Gasaway, Christof Iserlohe, Evan Kress, Ken Magnone, Nick Magnone, Michael McElwain, Juleen Moon, Gunnar Skulason, Inseok Song, Michael Spencer, and Jason Weiss • Lenslet Array Integral Field Spectrograph – Dissects arcsecond sized regions of the sky in 2 dimensions • • • • Spectral resolution sufficient to take advantage of low background between OH sky lines (R=3900) Full z, J, H, or K spectra with single exposure (1700 pixels) Very sensitive due to the suppression of atmospheric emission lines, the lack of slit losses and the low noise detector. Size ~ 1.5 tons – > – Vacuum chamber is ~ 1 m3 – About 200 kg is taken to 70 K y l x Design Summary (from Larkin) Lenslet Array Cold Pupil AO Focus Spectrograph Collimator Optics Grating Filters R. I. Collimating Singlet R.I. Camera Singlet Pupil Plane Focal Plane Lenslet Reimaging Optics Camera Optics Detector Lenslet Array Pupil Plane AO Focus (from Larkin) 1mm • MEMs Optical’s design is fused silica, biconvex elements. Thickness is 1.0 mm with EFL of 0.8 mm. Pitch is 250 microns. – 72x72 lenslet square area centered in 1.5” diameter circular substrate. • 2-3 microns of rounding between elements (98% fill factor) • 2 micron alignment front to back • Sub-micron accuracy of pitch • 1% variation in EFL across array. Example of microlens array - for Supernova factory integral field spectrograph (SNIFS) A symmetric magnifier for the spectrograph Insertable convex mirror Imager mirror Dark hole Beam from coronagraph Imager field Outer field angles To spectrograph Microlens array Microlens element of array for IFS Microlens (eg: 250μ dia, f/4) From magnification stage Microlens-based Integral Field Spectrograph layout Precede by magnification stage Microlens (element of array) Insert for modes 2 and 3 Insert for mode 3 Focal plane Mask – insert for mode 2 Change: Field lens now precedes microlens Insert for mode 1 TPF prime planet/speckle spectral IFS data format - at microlens focus, and projected onto detector, without disperser FWHM of diffraction limit Microlens Spot at focus of microlens 200 x 200 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector TPF prime planet/speckle spectral IFS data format - without disperser, showing detector pixel spacing FWHM of diffraction limit Microlens Spot at focus of microlens Detector pixel row spacing 200 x 200 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector TPF prime planet/speckle spectral IFS data format - with disperser 200 x 200 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector TPF prime planet/speckle spectral IFS data format - with prism disperser TPF baseline, 100 pixels for R~70 200 x 200 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector TPF prime and auxiliary candidate spectrographic capabilities Suggested point design layouts for science discussion and prioritization. 0.5 -1.0 micron range. 4k x 4k detector. 6.5 m telescope. TPF Baseline Mode 1 Mode 2 Mode 3 Resolving power. Pix/spectrum ~70 100 200 – 20000 100 3000 4000 3000 4000 Spatial Image points Ang resolution (arcs) Ang range (arcs) 200 x 200 0.037 x 0.037 3.7 x 3.7 200 x 200 0.037 x 0.037 3.7 x 3.7 200 x 5 0.037 x 0.037 3.7 x 0.185 33 x 33 0.22 x 0.22 3.7 x 3.7 Disperser Prism Grating (tiltable). + Filter Grating Grating Ensure other grating orders avoid detector Microlens array Concentration factor # interleaved spectra f/#s 200 x 200 20 5 500:4 200 x 200 20 5 500:4 200 x 5 (mask?) 20 5 500:4 33 x 33 120 30 500:4 Exchange microlens array for mode 4 Science capabilities Planets/speckles Circumstellar disks. Galaxy dynamics Galaxy jets Stellar jets SN in galaxy QSO in galaxy Galaxy pop synthesis. Gravity lenses? High z galaxy mergers Stellar ejecta (PN etc) Comments General astrophysics short high resolution spectral IFS data format - with grating disperser and filter range blocker Mode 1, 100 pixels for R~200 – 20,000, for selected short spectral regions 200 x 200 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector General astrophysics long medium resolution spectral IFS data format - with grating disperser Mode 2, 4000 pixels for R~3000, for long spectra Spatial format 200 x 5 microlens array. 20 pix separation. 5 spectra interleaved, each separated by 4 pix. 4k x 4k detector IFS microlens concentration factors Concentration factor limitation contributors: diffraction, projection, chromatic focus change, monochromatic aberrations. Diffraction: The FWHM at the focus of the microlens is given by dD=fλ/D, …. where D= lens diameter, f=focal length. Concentration factor, CD = D/d = D2/fλ …….for CD large, lenses should be large and fast. (For OAO 250μ f/4, CD = 62, dD = 4µ) Projection: Concentration factor, CP = D/dp = D/fφ = θ/φ, ….where f/#in = φ, and f/#out = θ (For a 10-m dia telescope (TPF), for a 250μ dia microlens to Nyquist sample the diffraction limit diameter of 0.020 arcsec at λ = 0.5µ, need f/# = 500. Then CP = 500/4 = 125.) Chromatic focus change: Concentration factor CF= D/dc For 250µ lens, from ray trace, best focus spot diameter = 2.4µ. Then CF = 100 Aberration: From ray trace of spherical lens, best focus spot diameter = 4.0µ. Then CA = 62 Combined concentration factor: C = √(1/ΣCi2) = 38, for the example above (250μ f/4 lenslets with f/500 input). Then with 4 pixel spacing between spectra, can fit 9 spectra between image points. Alternative Reflective Image Slicer Integral Field Spectrograph Detector Slicer – focussing elements Entrance slits Slicer – focal plane Slicer Difficulties • In a slicer spectrograph, one spatial axis (along the slice) is not sampled until the detector, so all optical components including the grating introduce noncommon path errors. • Each slice has different noncommon path errors. • Introduced polarization is also different for each slice. From: Content, 98 Example Image Slicer detector format - for 4 rows of re-imaging mirrors Adjusting for the Elliptical Primary - an asymmetric magnifier for the spectrograph Asymmetric magnifier to sample 8 m x 3.5 m diffraction limit at 0.5 microns Fixed convex cylindrical mirror, horizontally diverging Beam from coronagraph Insertable convex cylindrical mirror, vertically diverging Dark hole Fold mirror Outer field angles Microlens array Imager mirror Dark hole To spectrograph Spectrograph field Imager field Planet Detection Simulations (from Lindler and Heap, GSFC, at Caltech TPF-C meeting June 2004) • Four visits/two rolls per target • Exposure times set to detect Earth-sized planet at the scaled Mars distance. (limited to a maximum of 14 days for all visits) • Monte-Carlo simulation with random circular orbits he = 0.1 • PSF reset every 10,000 seconds to reduce speckle residual • Plots color coded by stellar luminosity • Spectra integrated over entire wavelength range to compute S/N 6 x 3.5 Imaging Mode: Readnoise=3, Darkrate=0.001 6 x 3.5 IFU Rpower=10 Readnoise=3 Darkrate=0.001 6 x 3.5 IFU Rpower=10 Readnoise=1 Darkrate=0.001 Planet Detection via Imaging vs. R=10 Spectroscopy Imaging • Standard CCD’s can be used • Requires Telescope roll to remove speckles • Requires a very stable PSF Spectroscopy • Requires detector with readnoise < 1e• No telescope dithers needed • Planet confirmation information obtained • Reduced PSF stability required