Direct Planet Imaging with a 30-m GSMT

Response to Astro2010 Programs Subcommittee Request for Information

Mitchell Troy Jet Propulsion Laboratory, California Institute of Technology

mitchell.troy@jpl.nasa.gov 818-354-4730

Instrument Team Leads

James R. Graham (UCB), Rene Doyon (University of Montreal), Motohide Tamura (NOAJ), Mitchell Troy (JPL/Caltech)

1. Summary The study of exoplanets is in an era of exploration a stage that will continue through 2020. In this upcoming decade, direct detection—spatially separating planetary light from that of its parent star—will become increasingly important. Early results (Figure 1) illustrate the power of this technique. With the advent of optimized AO coronagraph systems, e.g., the Gemini Planet Imager (GPI) and SPHERE (both with first light in 2011), hundreds of Jovian planets will be detected. Powerful though these instruments are, they will primarily image relatively massive (1-10 Jupiter masses, MJ), moderately young (10-1000 Myr) planets at angular scales of > 0.1 arcseconds.

Figure 1: Left: Coronagraphic images showing the exoplanet Fomalhaut b (Kalas et al. 2008). Inset shows the orbital motion from 2004 to 2006. Right: Three massive planets orbiting HR8799 (Marois et al. 2008).

As a diffraction-limited process, planet imaging scales as Dtel4, where Dtel is the telescope

diameter. In the next decade, Extremely Large Telescopes (ELTs) will unlock new exoplanet science, including high-SNR/high-resolution spectroscopy, characterization of mature exoplanets through reflected light, a complete census of exoplanets from the Doppler period cutoff to exo-Kuiper belts, small inner working angles that enable imaging on planet formation scales (3-5 AU) in local star forming regions, and the possibility of detecting and characterizing water-world “super-earths”. This capability was recognized by the AAAC Exoplanet Task Force as the major long-term ground-based priority (Lunine et al. 2008.) The GSMT observatories have recognized the importance of direct exoplanet imaging. However, this is a goal that cannot be achieved with typical instruments. The technological challenges of a high-contrast imager, and the attendant high cost, more naturally lead to a stand-alone project that complements the Observatory's more general capabilities. Such an instrument is a fully-fledged science experiment that would effectively leverage the large international investment in GSMT. Investment by a national organization is required, either NSF or NASA (supporting scientific goals and technology for space coronagraphs). The purpose of this white paper is to establish the technological feasibility, scientific capabilities, and approximate cost of such an instrument. We take as a baseline the Planet Formation Imager (PFI), a powerful AO coronagraphic imaging spectrograph designed for the Thirty Meter Telescope (Macintosh et al. 2006). However, the basic scientific capabilities are appropriate to any ELT—the key message is not that this specific system should be built, but that some capability be planned for whichever GSMT is constructed, and also that (equally importantly) no design decisions in a GSMT preclude this class of science.

2

2. Science motivation In the next decade, reflex motion, transits, and microlensing will be used to discover terrestrial planets. These techniques, however, will only measure orbital parameters, masses, and radii in even the best cases. The exploration of the physics of planet formation, structure, and evolution will focus on giant planets (M > 10M⊕ with hydrogen envelopes.) We discuss briefly here the science that direct detection and spectroscopic characterization of giant planets by an ELT will enable in the study period; see the white paper by Macintosh et al. (2009) for more details. Key priorities for exoplanets in this epoch are:

• Accumulation of exoplanet statistics across host-star spectral types (BAFGKM) to enable quantitative tests of the formation and evolution of solar systems stars, to understand how mass and luminosity of the host star impacts planet formation.

• To explore solar systems beyond the frost line. • Complete sets of orbital elements to investigate the dynamical state of exoplanets. • Detection and spectroscopy of exoplanet light to study their atmospheres, measure

effective temperature, gravity, and composition. • Reconstruction of the thermal history of Jovian planets to explore the early stages of

formation, including studying the formation of planetary systems in situ. • Explore and delineate the relationship between exoplanets and debris disks.

Realizing these will require significant investments in dedicated instruments and theoretical capabilities. This paper focuses on the scientific capability needed to enable detailed studies of giant planets—as astrophysical objects, rather than orbital elements: a dedicated high-contrast imaging system on a 30-m-class ELT.

2.1. Capabilities of direct imaging High contrast imaging systems can be characterized by an inner working angle (IWA), the smallest offset at which a planet can be detected, and a contrast at which a source can be reliably seen. For example, Jupiter is ~ 10-9 fainter than the sun; at 20 pc the separation would be 0.25”. Typical IWAs will be 2-5 λ/Dtel. Near-future systems on 8-m telescopes should achieve near-IR contrasts of 10-7 at 0.15”. This allows detection of self-luminous giant planets through their retained heat of formation at ages of up to 1 Gyr for massive planets (Figure 2). Similar instruments on ELT may exceed contrast of 10-8 with IWA of 0.04” (Figure 3). At this level, almost any self-luminous planet can be characterized at high SNR and spectral resolution, and mature planets (including Doppler exoplanets) in the inner parts of solar systems (< 2 AU) become detectable through reflected light. Dedicated space coronagraphs may achieve visible-light contrast of 10-9 or even 10-10, though their smaller aperture will limit IWA to 0.2” even at short wavelengths.

3

Jupiter

T dwarfs

Figure 2: Surface gravity/Teff for exoplanets. Hot, young jupiters detectable with near-future AO systems lie at the center of this figure—solid dots can be detected by the Gemini Planet Imager. Solid curves are evolutionary tracks; dotted curves are isochrones (Burrows et al. 2003). Condensation curves for H2O and NH3 are dashed lines. The coolest known T dwarfs are in the upper right (Burnigham et al. 2008; Burgasser et al. 2006; Warren et al. 2007). Mature, solar-system planets (4.5 Gyr) are located at the lower left. Space coronagraphs or GSMT AO will be capable of detecting mature planets from 0.3 MJ or lower filling the 100-800K regime. The planets characterized by transit techniques occupy a narrow region off the far right of the graph.

1.5-m Space coronagraph

Figure 3: Contrast vs. separation for notional exoplanet populations, showing the terrestrial planets (lower left), Jovian planets (black: old jovian planets seen in reflected light; green: younger self-luminous planets) and accreting protoplanets in Taurus (top left). Contrast curves are for an 8-m (Gemini Planet Imager), a 30-m telescope ExAO systems (at 1.65 μm), and a high-performance 1.5-m space coronagraph (at 0.8 μm).

2.2. Exoplanet statistics beyond Doppler searches Although there is preliminary information regarding the statistics of mass and semimajor axis of exoplanets, the current catalogs suffer from incompleteness for masses below 0.3 MJ and periods longer than 5.5 years (3 AU). About 5% of targeted stars reveal Doppler-detected exoplanets: the

4

abundance of circumstellar disks suggests that the frequency of planetary systems may be 15–50% and the low detection rate of planets may be a consequence of the biases inherent to detection of orbital motion. By 2020 Doppler searches will have reached 8 AU and direct searches with ELTs will explore from 1-40 AU, producing a complete picture. Characterizing the frequency and orbital geometries of planets beyond 3-5 AU will enable us to answer whether configurations like our own planetary system are commonplace. Direct detection will also reveal the zone where planets may form by direct gravitational instability and uncover traces of planetary migration, and open up spectral types (A and early F) and ages (< 1 Gyr) that are poorly suited to Doppler techniques.

2.3. Planetary characterization Direct detection offers the opportunity to characterize planets at large orbital separation and opens an important new window into understanding the process of planetary formation and evolution in other stellar systems. Characterization of giant exoplanets begins with constraining mass, radius, and bulk composition and moves on to measuring atmospheric chemistry and recognizing important global processes, such as cloud condensation, stratospheric heating, atmospheric dynamics, and photochemistry. By studying planetary architectures and processes, direct imaging and spectroscopy brings studies of extrasolar planets into the realm of comparative planetology, as has been practiced within the solar system over the past sixty years.

1 2 3

Figure 4: Luminosity versus time for a 2 MJ planet (Marley et al. 2007). The thick solid curve includes the effects of core accretion-gas capture. The planet is fully formed at 2.2 Myr. The dashed curve shows the simple cooling track of early “hot start” models. The full-width at half-maximum of the accretion luminosity spike is ~ 40,000 years. In reality, the spike is likely to be broader because of gradual accretion across the gap that the protoplanet forms. Numbers refer to the three phases: 1) solid accretion; 2) hydrodynamics gas accretion; and 3) run away gas accretion.

The most fundamental measure of a planet is mass. This can be estimated for luminous young planets by comparison with evolutionary models, e.g., the HR 8799 system. The properties of young planets depend upon the conditions established during their formation and thus such mass determinations are currently uncertain. The standard theory for the formation of gas giant planets is the core-accretion model (e.g., Pollack et al. 1996), which begins with dust particles forming icy and rocky planetary cores. If the core becomes massive enough while gas remains in

5

the disk, it continues to grow by gravitational accretion of this gas. Further accretion is slowed by the dwindling supply of local raw materials and by the extended envelope, leading to growth times of 5–10 Myr. Marley et al. (2007) and Fortney et al. (2008) have conducted preliminary calculations that describe the cooling and contraction of a young planet as it emerges from its disc (Figure 4). The implication of these results is that giant planets formed by the core accretion-gas capture mechanism are less luminous post-accretion than had been previously anticipated. There are two significant observational consequences: 1) there is a period of very high luminosity (likely broader and fainter than in the idealized calculations above); 2) the initial conditions for evolution models are not “forgotten” for tens of millions of years. These factors imply that observations of planets with ages 0-100 Myr, particularly if their mass can be independently determined, afford the opportunity to probe the planet formation event in ways that distinguish between different formation scenarios. As more planets are imaged around stars of known ages, our knowledge of planet formation will improve and the systematic errors will be reduced. Direct detection of planets also detected by radial velocity or astrometric methods or with masses determined from mutual interactions or other dynamical methods (e.g., the mass of Fomalhaut b is constrained by its perturbation of the debris disk) will ultimately serve to validate planet formation models. A wide gap in Teff and log g exists between the currently known T dwarfs and cool, solar Jovian planets (Figure 2). However, these objects must exist as the youthful progenitors of the known population of Doppler-detected exoplanets. Figure 5 shows theoretical spectra of a 5 MJ exoplanet as a function of age showing the distinctive peaks due to enhanced flux between the water vapor absorption bands. Thus, the ground-based near-IR YJHK bands, which are defined by the same H2O opacity, are ideal bands in which to seek detection. Spectroscopy of planets opens their atmospheres to the study of temperatures, gravities and compositions. These objects represent planetary terra incognita. For example, at Teff's below 400–500 K water condenses in planetary atmospheres; a second major transition is expected to occur below 180 K when NH3 clouds form. The appearance of water ice clouds constitutes a significant milestone along the path from the known T dwarfs to the giant planets. The atmospheres of all of the solar system giants are enhanced in heavy elements over solar composition, by factors ranging from 4 to 50 for various elements and planets. Generally the enhancement increases with distance from the sun and with falling mass. The detailed pattern of the enhancement is taken to be a signature of the planet formation process and the subsequent accretion of planetesimals. The details of this interpretation in the solar system remain controversial. A measure of the atmospheric composition for many different planetary systems will thus provide important new tests for planetary formation theory. We are already on the cusp of such discovery with the three planets orbiting HR 8799. Even relatively low-resolution, near-infrared spectra of these planets will reveal their gross composition. The substantial work that has already been done to understand the well-studied brown dwarfs paves the way for such studies and validates much of the spectral modeling approaches. The presence or absence of cloud layers, which can also be discerned from low-resolution spectra, also serves as an atmospheric composition and temperature indicator.

6

Figure 5: Spectra of a 5 MJ exoplanet as a function of age showing the distinctive peaks due to enhanced flux between the H2O bands (0.93, 1.1, 1.4, 1.8 µm). Other features are the broad 4.5 µm hump, CH4 (1.7, 2.2, 3.3 µm) and NH3 (1.5, 1.95, 2.95 µm). The strengths of each of these features are functions of mass and age.

2.4. Imaging the process of planet formation PFI will achieve angular resolutions of 10 mas or better, with an inner working angle of 30 millarcseconds. This enables observation of the planet formation process itself in nearby (140 AU) young associations or star-forming regions. Bright, actively accreting planets may be directly detectable, particularly if they have opened gaps in their host disk. Even in optically thick disks, planet embryos in gas-rich disks can locally perturb the disk structure, resulting in shadows at the surface of the disk, which may be observable at high angular resolution. Figure 6 shows simulated scattered light images of planet cores of various masses and distances at 1 and 3 μm The spatial scale and contrast of the shadow increase with planet size and approximately linearly with the distance between the planet and the star.

Figure 6: Recent scattered polarized light observations of the Herbig Ae star AB Aur show evidence of a shadow in the disk that could indicate the presence of a planet (Oppenheimer, et al. 2008). Scattered light images of planet cores of various masses and distances (Jang-Condell, in prep).

7

3. Technical overview Fully realizing this scientific potential will require not general-purpose instruments, but rather dedicated high-contrast systems installed on an ELT. High-contrast imaging with current telescopes and AO systems is limited by several factors:

• Large (40-80 cm) AO subapertures; • Aliasing of high-spatial frequency wavefront errors, both atmospheric and telescope-

induced, in wavefront sensing (Poyneer et al. 2004); • Miscalibration of the non-common-path optical errors between the wavefront sensor

and science instrument; • Surface errors on internal optics; • Diffraction from the telescope primary mirror. Coronagraphs can mitigate this, but at

low Strehl ratio, most coronagraphs are ineffective. Advanced image processing techniques, or differential imaging, can partially mitigate these effects, but not entirely, especially at small separations. Figure 7 shows typical contrasts for current systems as well as the predicted performance for GPI.

Figure 7: Contrast comparison for current AO systems, from the Gemini Deep Planet Survey (LaFreniere et al. 2007), Keck adaptive optics (e.g. Marois et al. 2008), and the Gemini NICI dual-channel imager (Chun et al. 2008.)

These effects can be overcome in a dedicated high-contrast system (sometimes referred to as “Extreme” adaptive optics, or ExAO). The Gemini Planet Imager (Macintosh et al. 2008) is an example currently under construction for the Gemini Observatory. Key features for such systems include:

• High-density AO (1800 actuators) for Strehl ratio of ~0.9 at H at I = 8 mag.; • Spatially filtered wavefront sensors to prevent aliasing; • Removal of quasi-static non-common-path errors through a dedicated (slow) infrared

wavefront sensor integrated with the coronagraph; • High-quality optics (~2-4 nm RMS wavefront error); • High-performance coronagraph tuned for ground-based telescope apertures; • An imaging integral field spectrograph optimized for planet detection and

characterization (R ~ 50, 0.014 arc second pixels, 2.8 x 2.8 arc second FOV) and polarimetric capability.

8

Detailed near-field diffraction simulations and integrated modeling of GPI indicate that it should reach contrasts of 10-7. GPI has passed Critical Design Review and is on track for first light in early 2011. With the lessons learned from GPI, and scaling to a 30-m telescope, contrast of 10-8 or better, at much smaller inner working angles, should be practical.

3.1. Planet Formation Imager The TMT Planet Formation Imager (PFI) is the result of a feasibility study for a high-contrast system for TMT (Macintosh et al. 2006). We report on the key requirements and features of the design in the following sections. Building on the GPI design, PFI is optimized for the science goals discussed above – which require contrast on the order of 10-8 at inner working angles as small as 30 milliarcseconds, ~3 λ/D for a 30-m telescope.

• A pyramid wavefront sensor to allow better AO performance down to I = 10 mag. • A high-speed interferometric IR wavefront sensor, integrated with the coronagraph. • Multiple deformable mirrors to allow control of both phase and amplitude errors. • A shearing interferometer coronagraph designed for very small IWA. • Integral field spectrograph with R = 70 and R = 700 modes.

Figure 8 shows a block diagram of the PFI system including: the front-end AO system consisting of a DM, WFS1 and Controller 1 (Section 3.2), the DSS and IR WFS consisting of the DSS, WFS2 and Controller 2 (Section 3.3) and the science IFU (Section 3.4).

Property Requirement Goal PFI value Inner Working Angle (IWA) 0.03 arc sec 0.03 arc sec @

H band Contrast (I < 8 mag.) @ IWA 10-8 @ 50 mas 2x10-8 without

speckle suppression

Contrast (I < 8 mag.) wide angle 10-9@ 100 mas Contrast (H<10 mag.) @ IWA 10-6

@ 30 mas 2 10-7@ 30 mas 1x10-6 without speckle suppression

Guide star limit H<10 mag. H<11 mag. H<11 (H<12 possible with suitable IR detectors)

Plate scale Nyquist @ H Nyquist @ J 5.5 mas Field of View (radius) 0.7 arc sec 2 arc sec 2 × 2 arc sec Spectral resolution, full FOV 50 100 70 Spectral resolution, partial FOV 500 1000 700 Wavelength range 1–2.5 μm 1–4 μm 1–5 μm Imaging polarimetry Simultaneous

two channel SDC

Sensitivity (1 hr., 5-σ) H = 27 mag. H = 32 mag. H = 32 mag. Table 1: PFI Requirements. Speckle suppression processing of IFS data cubes is expected to increase contrast by a factor of ~ 10 for methane-dominated planets.

9

Figure 8 Simplified block diagram of the PFI system.

3.2. Adaptive optics Like GPI, PFI is designed around MEMS deformable mirrors—128x128 actuators in the baseline. These will require technological development (see Section 4 for more details). A cascade of 2 DMs (one not at a pupil) provides broadband control of both phase and amplitude errors. Current-generation ExAO systems use Shack-Hartmann wavefront sensors. These are proven technology able to measure even uncorrected atmosphere wavefronts, but they are very inefficient for measuring corrected wavefronts. Several classes of interferometric wavefront sensor can measure the crucial low spatial frequencies with sensitivity an order of magnitude better. For PFI, we selected the pyramid wavefront sensor operated in an interferometric (non-dithered) mode and expect, with predictive control (Poyneer et al. 2007) that the visible-light WFS will operate with near-full performance down to I = 11 mag., compared to I = 8 magnitudes for GPI.

3.3. Diffraction control and IR WFS PFI requires a diffraction control system (coronagraph) with a very small inner working angle. As a baseline, we selected the “visible nuller” (Samuele, R. et al. 2007). This combines two or more sheared and phase-shifted copies of the pupil to cancel the coherent part of the wavefront. Simulations indicate this can obtain an inner working angle of 2-3 λ/D at PFI contrast levels with an effective throughput of ~25%. A key feature of the GPI architecture is the use of an interferometric post-coronagraph wavefront sensor that measures the wavefront at the science wavelength and coronagraph location. The GPI version of this sensor operates at ~1 Hz rates, measuring time-averaged wavefronts to look for small static biases. In PFI, this IR sensor uses interference between the science light (from the “dark” output of the nulling interferometer) and the bright rejected light. It is designed to operate at ~1 kHz frame rates. In addition to measuring static errors, it can now measure (and correct with a dedicated DM) dynamic atmosphere errors. On bright stars, cascading the visible and IR AO systems provides much steeper rejection of low temporal frequency errors. On dim, red

10

targets, such as embedded young stellar objects, the IR WFS becomes the primary correction source, allowing PFI to operate on targets as dim as H = 11 mag. with I-H < 4 mag.

Stop

Starlight frompre- AO

Mach-ZenderNuller (DSS)

DeformableMirror

Modulator

DM Controller

WFS Camera

Processor

Science IFU

SpatialFilter

Figure 9 A functional schematic of the backend AO system.

Figure 9 shows a schematic of the backend AO system and DSS. A shearing MZ Nuller provides a nulled and a bright output to the backend. The nulled output is split 50:50 between the Science Camera and the post-DSS AO. The bright output is spatially filtered, modulated and interfered with the nulled output at the main beamsplitter. In this case, the combined pupil emergent on both sides of the beamsplitter is reimaged onto a focal plane array. The processor acquires the WFS data and generates a phase map error, which is handed of a controller to close the loop with a DM, which is placed in one arm of the nuller.

3.4. Science instrument Like GPI, the primary science instrument is an imaging integral field spectrograph. This dissects the field of view using a lenslet grid and then disperses each sub-image. This design is high throughput and induces no wavelength-dependent PSF errors – so that the resulting spectral cube can be used to distinguish true companions from PSF artifacts (Marois et al. 2006). For PFI, the spectrograph is designed to operate out to 4.8 μm, allowing observations of highly obscured planets in protoplanetary dust disks and dust/ice spectral features. The instrument supports both a R = 70 mode for initial planet discovery and a R = 500-700 mode for characterization. Finally, a dual-channel polarimetry mode can be used to study circumstellar dust (including extrasolar Zodiacal dust at 1-5 AU scales).

3.5. Effects of the telescope design and implementation Unlike space coronagraphs, ground-based ExAO systems are implemented on existing general-purpose telescopes, which may have complex apertures. During the PFI design study we examined the effects of the telescope on instrument performance (Troy et al. 2006).

11

One crucially important result to note is that a highly segmented aperture (such as the TMT design) has little effect on high-contrast performance. In fact, a large number of small gaps are much better than a small number of large gaps. This can be understood quantitatively with a simple analysis. Consider a telescope of diameter D with segment size d and gap size g; N = D/d is the total number of segments across the aperture. The gaps will scatter light into a series of sidelobes located at multiples of λ/d = Νλ/D. The larger the number of segments, the further the first sidelobe will be from the scientifically crucial region of 2-5 λ/D. For g<<d, it can also be shown that the total amount of power in the first sidelobe scales as g2/d2. (i.e., proportional to the fill factor of the gaps.) Small gaps produce a large pattern of relatively diffuse sidelobes; and small segments locate those sidelobes far away from the target star. The TMT design (large N, small d and g) is significantly better than the GMT design (small N, large d and g) in this respect. To the extent that diffraction from segment gaps is an issue, it can of course be mitigated by a suitable coronagraph, but any coronagraph imposes penalties in terms of throughput, wavefront error sensitivity, etc; the less severe the initial sidelobes are, the simpler and more effective the instrument will be. The analysis above indicates that larger structures can have worse effects. Most GSMT designs have quite large supports for their secondary mirror, e.g., 40 cm for the TMT design. These must be dealt with by the diffraction suppression system. The shearing nulling interferometer has the advantage of a producing a very sharp output pupil, in which only the secondary supports themselves need to be masked rather than an oversized region. Another concern is reflectivity variations as individual segments are continuously being removed for re-coating and replaced with fresh segments; reflectivity will vary randomly across the primary mirror. Simulations indicate that this reflectivity variation would have to be kept significantly below 1%. Since this is impractical, PFI will incorporate multiple deformable mirrors to allow correction of amplitude errors (Shaklan et al. 2006) over the full field of view and broad (20%) bandpass. We have investigated the impact of phase errors for the TMT design on contrast. The M2 and M3 phase errors can easily be corrected by the AO system. The largest impact to contrast comes from M1, but not from piston/tip/tilt phasing of segments (~13 nm RMS wavefront), but rather residual segment aberrations (~25 nm RMS. wavefront) End-to-end simulations of the baseline PFI design indicate the 5-sigma contrast contribution from the telescope after AO and coronagraph will be ~10-7 from 3-10λ/D. These errors are relatively static and achromatic, and will be removable via multiwavelength or temporal processing of images. A complete set of detailed physical-optics simulations (including near-field diffraction and chromaticity effects that will limit post-processing) needs to be performed for any GSMT design to insure the telescope does not prevent an instrument from reaching the required contrast levels.

4. Technology drivers There are three areas in which technology development is needed in order to reach the contrast levels listed in Table 1.

12

Diffraction Suppression Systems (DSS) As described above a diffraction suppression system (often referred to as a coronagraph) removes the coherent part of the scattered light (i.e. the Airy pattern) from a bright star. We require the suppression of the photons from the host star to levels on the order of 10-6 to 10-9 at angles as small as 2-3 λ/D. There is significant research, including testbed verification, going into DSS for space based telescopes, but these are not applicable to ground base telescopes because: 1) they are designed for un-obscured telescopes and 2) they typically have a high sensitivity to residual image motion. The obscuration from the secondary mirror and to a significantly lesser extent the segmentation of ground-based telescopes creates a significant problem for most DSS. Several designs and concepts exist for DSS suitable to ground-based work, including interferometric nullers, apodized-pupil Lyot coronagraphs, and modifications of the phase-induced amplitude apodization. As described above the baseline DSS for PFI is nuller, which in theory can reach very high contrast levels, however, the required levels have not yet been demonstrated in laboratory testbeds. Near-term instruments such as the Gemini Planet Imager will demonstrate these at the 10-7 @ 5 λ/D level, but significant analytical (including integrated models), laboratory and on-sky testing are needed to reach the levels of performance required for future ELTs.

Advanced wavefront sensors The majority of AO systems use Shack-Hartmann or curvature wavefront sensors (WFS). These are robust, with high dynamic range, but inefficient—it has been shown (Guyon, 2005) that as currently implemented these are insufficient for high-contrast imaging due to the relatively poor sensitivity to low spatial frequency aberrations. Several new techniques for wavefront sensing that exploit the coherence of a high-Strehl ratio ExAO correction, such as a Pyramid sensor (operated in a non-dithered mode), the Zernike phase contrast sensor, Mach-Zehnder interferometer, and focal plane interferometric WFS have been proposed and analyzed. In principal, these could improve wavefront sensing SNR by a factor of 4 or more. However, in the majority of these cases these sensors have not been built and tested in a laboratory or on the sky. It is critical that these sensors are built, characterized in the lab and then tested in field systems to gain an understanding of how they really work.

Deformable mirrors PFI will require several (three) deformable mirrors each with ~128 X 128 actuators. Such devices do not currently exist, however, both Xinetics and Boston Micromachines have or are close to having 64 X 64 actuator devices. Xinetics uses a macro-scale technique producing DM’s with actuator pitch’s ranging from 0.5 mm to 7 mm and corresponding actuator stroke ranging from ~0.5 to 5 microns. Boston uses a MEMS technique with actuator pitch’s of ~300 microns and achieves a stroke of 4.0 microns. Neither device has sufficient stroke to correct the atmosphere and a 2nd “woofer” DM is needed.

Scaling these devices to 128 X 128 actuators will not be trivial. The first problem, which is common to both device technologies, is dealing with the needed cabling. If each actuator were cabled, then there would be 12,000 wires needed. Clearly some sort of high-speed multiplexer is needed, note these DMs need to support update rates on the order of 2000 Hz. ASICs mounted with the deformable mirror, or CMOS multiplexers bonded directly to a MEMS (similar to a hybrid IR detector array), are one possible approach, but significant R&D is required. Integrated

13

multiplexing will significantly simplify future adaptive optics systems, enabling not just ExAO but compact low-cost AO for general purposes ranging from small telescopes to multiple deployable AO IFUs (MOAO).

5. Organization and current status In 2005-2006, a team carried out a detailed feasibility study for the PFI system. Primary partners in this initial study were the Jet Propulsion Laboratory (PI: Mitchell Troy), Lawrence Livermore National Laboratory (PI: Bruce Macintosh), University of Montreal (PI: Rene Doyon), and University of California Berkeley (Project Scientist: James Graham.) Actual construction of the instrument would involve a larger collaboration, and we have had discussions with additional partners for large-scale optomechanical work, including NAOJ, HIA, and UC Observatories, as well as commercial aerospace firms The feasibility study identified a buildable design that can meet the scientific requirements. The TMT organization strongly endorsed PFI as a crucial scientific capability for the observatory. However, baseline TMT funding supports only three (at most) initial science instruments, and these limited resources are directed towards workhorse capabilities such as optical spectroscopy or general-purpose AO. The TMT project remains committed to a high-contrast imaging capability, and PFI team members are involved in discussions about telescope design to insure that planet imaging is not precluded by the final telescope. PFI is still on the list of desired “first light” instruments, but no funding source has been identified for it. Building such a capability for a GSMT project will almost certainly require a dedicated funding commitment by a national agency, either NSF or NASA. High-contrast imaging was the highest long-term ground-based exoplanet study priority identified by the AAAC Exoplanet Task Force, but translating that into an instrument will be a complex process. The project price is too high for current NSF MRI funding (as will be true for any facility instrument on an ELT) and observatories are unlikely to use TSIP funds (with associated telescope time cost) for a specialized facility. Co-funding by the observatory and a private donor are a possibility, but a major commitment by a federal agency (or international partner) will likely be needed as well. The Decadal Survey should evaluate paths for NSF funding of ELT instruments of all kinds, especially in the context of scientific priorities identified by reviews such as ExoPTF. Construction of such an instrument could be an outcome of participation by a national observatory (US or Japanese) in TMT. Another possibility would be NASA funding – PFI could accomplish almost all of the giant-planet scientific goals of a Explorer or Discovery-class space coronagraph, while also validating technology for a future flagship Terrestrial-planet search mission.

6. Activity schedule As part of the 2006 PFI feasibility study we developed a schedule for building the instrument. The schedule was based on the detailed proposal for building GPI as well as experience with AO systems at Palomar and Lick. The schedule presented below has been updated with the experience gained in actually developing and building GPI. Unlike with GPI there is also the need for some basic technology development, which is also called out in the schedule.

14

The schedule is organized in the following programmatic phases: • Technology Development – Develop the technology items listed in Section 4. This would

include analytical, laboratory and on-sky testing. This also includes working with the selected GSMT telescope to insure that the telescope meets the optical quality requirements needed for high-contrast imaging.

• Conceptual Design – Final requirements definition, architecture trade studies, and production of an updated cost estimate and schedule for building the instrument.

• Preliminary Design – This is the standard preliminary design phase, which ends with a preliminary design review.

• Final Design – This is the standard final (or critical) design phase and ends with a final design review.

• Build and Test Subsystems – Each sub-system will be built and tested as a unit. This phase ends with unit acceptance testing and pre-ship review.

• Integrate – In this phase we integrate the subsystems and perform the final integrated performance testing to demonstrate the system can met the requirements. This phase ends with a pre-ship review.

• Install and Commission instrument at the telescope – In this phase we install the instrument on the telescope and perform commissioning, which will include performance optimizations. This phase ends with start of science operations with the instrument.

• Dedicated Science Campaign – A dedicated science observation campaign will be conducted for the three years.

Figure 10 shows a top-level schedule for the project with first light occurring approximately 1 year after first light for TMT; we presume GMT has a similar schedule. The start of the conceptual design phase in FY14 will allow us to incorporate not only what we have learned from the technology development phase, but also knowledge gained from GPI, SPHERE, and PALM-3000 as well. The elapsed time from start of conceptual design to being ready for science is 6 years, which is consistent with other large instrumentation projects. The technology development phase is critical for this project. Without a demonstration of the needed technology before the conceptual design phase starts the cost and schedule for building the instrument has large uncertainties. We estimate that in three years we can develop and demonstrate the needed technology to make this instrument a success. This includes demonstration of a 128 X 128 DM, on-sky validation of at least two advanced wavefront sensing techniques and laboratory validation of at least three DSS techniques. The end result of this technology development is that we will have reduced or eliminated the major risks to the project.

15

Figure 10 Top level project schedule

7. Cost estimate In this section we break the cost estimate down into three major categories: technology development, design and build the instrument and operations costs.

Technology development As part of this white paper effort we have developed a plan and cost estimate for the needed technology development. The DM technology development effort will develop the needed requirements for the DMs and through a competitive process fund two or more approximately 6 month in length at companies that are already manufacturing DMs. With the results of these studies in hand we will select one or two companies to go forward with an approximately 24-month contract and develop the needed 128 X 128 DM and electronics. Once we receive these devices we will test them in the laboratory to insure they met the specified requirements. Table 2 shows the estimated cost of $3.2M. It may be possible to fund some of this development via SBIR grants. The end result of this development will be a 128 X 128 DM that can meet our requirements as well as an associated cost for the purchase of future DMs. We will develop at least one advanced wavefront sensor that can meet the instrument requirements. We will start with 6 months of analysis and simulation to better understand which WFS are ideal for our particular application. A result of this study will be the selection of approximately 3 WFS ideas to build and perform laboratory tests with. Each WFS concept will be built, tested in the lab and the performance results compared against the results expected from simulations. Two WFS concepts will be selected and demonstrated on the sky in closed loop. This experiment would be carried out on an existing AO system, such as at Palomar, Lick or the MMT. This phase will include on-sky tests over a six-month period and will include analysis of the results. At the end of this phase we will have at least one advanced WFS that can meet our requirements. Table 2 shows the total estimated cost of $2.8M, which is also broken down by program phase. We will develop at least one DSS that can meet the instrument requirements. We will start with a year of analysis and simulation to further develop DSS that are optimized for the selected

16

GSMT. As a result of these studies we will select approximately 3 ideas to carry forward to laboratory testing. This laboratory testing would occur over a 2 year time period and would leverage the existing high-contrast testbeds such as: the high-contrast imaging testbed at JPL (built for TPF), the UCSC laboratory for AO testbed, and a NASA funded nulling testbed at JPL. These laboratory tests will be augmented to include testing with obscured and segmented apertures. As a result of this study we will have at least one advanced DSS design that will be demonstrated to achieve contrasts sufficient for this instrument (> 10-9). Table 2 shows the total estimated cost of $3.7M, which is also broken down by program phase.

Task Analysis & Simulation

Laboratory Tests

On-sky Tests

Procurements Totals

Deformable Mirror $40K $140K $3,000K $3,180K Advanced WFS $280K $1,240K $570K $750K $2,840K DSS $550K $1,650K $1,500K $3,700K Totals $870K $3,030K $570K $5,250K $9,710K

Table 2 Technology development effort cost estimate in fixed year FY09 dollars.

Design and build instrument The cost estimates summarized in this document do not constitute an implementation-cost commitment on the part of JPL or Caltech and have not been validated by JPL. As part of the 2006 PFI feasibility study a cost estimate was led by Lawrence Livermore National Laboratory. Each primary partner developed its estimates using their institutional processes. The estimates were produced in FY05 dollars and then inflated to FY09 dollars using inflation factors generated by TMT for similar purposes. The project was broken into 6 WBS elements (see Table 3). A cost estimate was generated for each WBS element and phase of the project. At this level labor was estimated based on similar projects including GPI, other instruments and testbeds. In addition procurement lists for the major components were generated for each subsystem. Table 3 shows that the total estimated cost to build this instrument is $46.2M, including 30% contingency. The figure does not include the cost for the needed technology development described in the preceding section.

Task CoD PD CD Build & Test

Integrate Install & Commission

Procurements Totals

System level Eng. & Mgt.

$300K $610K $610K $610K $610K $460K $3,200K

Pre-DSS AO $1,060K $1,830K $1,920K $2,290K $880K $330K $3,350K $11,670K Mechanical $140K $830K $1,130K $1,130K $1,020K $4,240K Post-DSS AO and DSS

$630K $950K $1,260K $1,580K $790K $240K $2,470K $7,910K

IFS $240K $400K $860K $680K $680K $120K $3,460K $6,450K

High-level Software

$140K $280K $540K $550K $280K $140K $120K $2,040K

Totals $2,520K $4,890K $6,330K $6,840K $3,240K $1,280K $10,420K $35,520K Contingency 30% $10,660K

Table 3 Estimate of cost to build the instrument in fixed year FY09 dollars.

17

Operations costs Based on the load of instrument specialists at other telescopes we estimate that the marginal cost for an observatory to support this instrument is 0.5 FTE over the life of the instrument (25 years). In addition we include a budget to support the previously described 3 year dedicated science campaign. There is also support for the first three years to continue the instrument performance optimization as the dedicated science program is being executed. The total estimated operations cost is shown in Table 4 and is $4.5M.

Duration

(Years) Level of support (People per yr)

Cost

Instrument specialist 25 0.5 $2,810K Continued system optimization 3 1.0 $680K Dedicated science program PI 3 1.0 $300K Postdoc 3 2.0 $450K Graduate student 3 2.0 $300K Total $4,540K

Table 4 Operations cost estimate in fixed year FY09 dollars.

18

19

Acknowledgments The authors would like to acknowledge that this document is based largely on the TMT PFI Feasibility study lead by Bruce Macintosh and the excellent work of that team consisting of the people listed below. This research was carried out in part at the Jet Propulsion Laboratory, California Institute of Technology, and was sponsored by the California Institute of Technology and the National Aeronautics and Space Administration. The authors gratefully acknowledge the support of the TMT partner institutions. They are the Association of Canadian Universities for Research in Astronomy (ACURA), the California Institute of Technology and the University of California. This work was supported as well by the Gordon and Betty Moore Foundation, the Canada Foundation for Innovation, the Ontario Ministry of Research and Innovation, the National Research Council of Canada, the Natural Sciences and Engineering Research Council of Canada, the British Columbia Knowledge Development Fund, the Association of Universities for Research in Astronomy (AURA) and the U.S. National Science Foundation. © 2009. All rights reserved.

Lawrence Livermore National Laboratory: Bruce Macintosh (PI) Kevin Baker Brian Bauman Victor Karpenko Christian Marois David Palmer Donald Phillion Lisa Poyneer COM DEV: Neil Rowlands Ken Tam Immervision Inc.: Simon Thibault University of California: Travis Barman James Graham (Project Scientist)

Jet Propulsion Laboratory: Mitchell Troy (Co-PI) Ian Crossfield Philip Dumont Joseph Green Marty Levine Bertrand Mennesson David Palacios Michael Shao Gene Serabyn Chris Shelton Gautam Vasisht James Wallace Universite de Montreal: Rene Doyon (Co-PI) Jean-Francois Lavigne Philippe Vallee

References Burgasser, A. J., et al. 2006, ApJ, 639, 1095 Burnigham, B., et al. 2008 MNRAS 391, 320 Burrows, A., et al. 2003 ApJ, 596, 587 Chun, M., et al. 2008 SPIE 7015, 70151V Fortney, J. J., et al. 2008, ApJ 683, 1104 Guyon, O., 2005 A&A 629, 592 Kalas, P. et al. 2008 Science 322, 1345 LaFreniere, D. et al. 2007 ApJ 670, 1367 Lunine, J. et al. 2008

http://www.nsf.gov/mps/ast/aaac/exoplanet_task_force/reports/exoptf_final_report.pdf

Marley, M., et al. 2007 Ap.J. 655, 541 Macintosh, B. et al. 2006

http://ao.jpl.nasa.gov/TMT/Publications/Technical/PFI_version_1_24.pdf

Macintosh, B. et al. 2008 http://www8.nationalacademies.org/astro2010/DetailFileDisplay.aspx?id=244

Macintosh, B. et al. 2008 SPIE 7015, 31 Marois, C, et al. 2006 A&A 641, 556 Marois, C. et al. 2008 Science 322, 1348 Oppenheimer, B., et al. 2008 Ap.J. 679, 1574 Pollack, J. B., et al. 1996, Icarus, 124, 62 Poyneer, L. A. & Macintosh B. 2004 JOSA 21, 810 Poyneer, L. A., et al. 2007 JOSA 24, 2645 Samuele, R., et al. 2007 IEEE 3, 1 Shaklan, S. B., Green, J. J. 2006 AO 45, 5143 Troy, M., et al. 2006 SPIE 6272, 62722C Warren, S., et al. 2007, MNRAS, 381, 1400