THE GIANT MAGELLAN TELESCOPE: 24 M APERTURE OPTIMIZED FOR ADAPTIVE OPTICS

Transcription

1 THE GIANT MAGELLAN TELESCOPE: 24 M APERTURE OPTIMIZED FOR ADAPTIVE OPTICS Roger Angel, Michael Lloyd-Hart and John Codona Steward Observatory, The University of Arizona, 933 N. Cherry Ave., Tucson, AZ Matt Johns Carnegie Observatories, 813 Santa Barbara St., Pasadena, CA ABSTRACT Construction has begun of the Giant Magellan Telescope, with the casting in July 2005 of the first of its seven 8.4 m circular mirror segments. Together these segments will form a single f/0.7 primary mirror with the collecting area of 21.5 m aperture and resolution at the diffraction limit of 24.4 m aperture. To aid in the implementation of adaptive optics for such resolution, the very large telescope is designed to be as stiff as possible against wind buffeting. It is also configured with a deformable Gregorian secondary mirror, segmented with seven 1.1 m circular segments conjugated to the primary segments. The secondary segments will be agile in tip, tilt and piston motions so that dynamic wavefront discontinuities from rigid body motion of the primary segments can be corrected, in addition to atmospheric distortion. Tip/tilt and piston corrections will be sensed from faint field stars and segment edge sensors. The initial AO correction modes will use the 4600-actuator deformable secondary alone. When observing faint fields, the detailed correction signal will be sensed with a multiple sodium beacon system integrated into the telescope, based on experience with the pioneering multi-beacon system at the MMT. For imaging at the diffraction limit, the beacon constellation diameter will be small and tomographic reconstruction will be used to obtain the wavefront distortion. For improved seeing over wider fields, the beacon and sensor array diameter will be increased to distinguish ground layer turbulence. Correction of this turbulence can be made over several arcminutes with the adaptive secondary, because it conjugates to the ground layer. The common optical wavefront sensor unit for both modes will be located ahead of the direct f/8.4 Gregorian focus, and fed by reflection from an 8 arcminute diameter tertiary dichroic mirror. The third initial AO mode will be high contrast imaging, with correction signals derived from star sensors at the pupil and focal planes. In this case the deformable secondary will be used both to correct seeing and, by additional shaping, to strongly suppress the diffracted star halo without the inefficiencies of a Lyot coronagraph. Contrasts of 10-7 at 0.05 arcsec and 2x10-8 at 0.1 arcsec are projected for integrations of a few hours in the H band. Instruments that exploit ground layer correction will be located at the direct Gregorian focus. Those operating at the diffraction limit will generally be located on a rotating platform above, fed by additional dichroic tertiary mirrors passing light through to the universal optical wavefront sensor. 1. INTRODUCTION Adaptive optics (AO) is especially advantageous for very large aperture telescopes because as aperture is increased stronger flux is combined with smaller diffraction limited images. It would hardly make sense to build any future giant telescope without AO. Larger apertures will require more correcting actuators, in proportion to area, and higher resolution wavefront sensors. But there is no fundamental limit to correcting larger apertures. Diffraction limited images of given Strehl ratio require the same photon flux per unit area in the guide star or beacon wavefront sensors, independent of aperture. Nevertheless, there are a number of new challenges to recovering the diffraction limit over very large aperture that must be addressed in the basic telescope design. Wind, for example, in larger structures will cause larger deflections and excite stronger vibrations, while the tolerance for wavefront aberrations, a small fraction of a wavelength, remains independent of size. Already some of today s 8 m aperture telescopes cannot be operated in wind, even at the seeing limit, without constant servo correction of image vibrations. For much larger telescopes, wind could easily induce larger and higher order wavefront distortions that could be comparable in strength to those of the atmosphere. In some respects, such as time-varying discontinuities at segment boundaries, wind induced distortion could be more troublesome. The adaptive optics system must cope with such discontinuities as well as providing for larger stroke to deal with atmospheric aberration that increases nearly linearly with aperture (depending on the outer scale of turbulence). 610

2 Laser beacons, essential to determine adaptive correction for faint objects, become also more difficult to use with very large aperture. The wavefront measured by large aperture of a single sodium beacon is no longer the same as that of a faint star in the same direction, because of the increased slant angles to the beacon (the cone effect or focus anisoplanatism). New techniques using multiple beacon constellations and tomographic analysis must thus be developed to determine stellar wavefronts, for correction of even small fields. Also, means must be found to deal with the reduced sensitivity of a beacon wavefront sensor for large apertures which resolve the scattering column in the mesospheric sodium layer. Fig. 1a. Overall view of the GMT, showing the laser beam path from the laser room at lower left. Fig. 1b. Star s view of the GMT. The sevensegment configuration of the primary mirror is reproduced in smaller scale at the secondary. These considerations have strongly influenced the design of the Giant Magellan Telescope (GMT), planned as a giant successor to the existing twin 6.5 m Magellan telescopes in Chile [1]. The partners in the telescope project are the University of Arizona, Carnegie Observatories, Harvard University, the University of Michigan, Massachusetts Institute of Technology, the Smithsonian Astrophysical Observatory, the University of Texas at Austin and Texas A&M University. To provide for the highest resistance to wind excitation, the telescope will use rigid glass primary mirror segments supported by a steel structure designed for the highest rigidity and low cross section to wind, and made extraordinarily compact by our adoption of an f/0.7 primary [2]. Wavefront discontinuities across the aperture will be minimized by use of very large segments to synthesize the primary mirror surface. Despite these features residual motions on short time scales (~ seconds and shorter) will likely be significant at the diffraction limit. To correct these, the telescope is designed about a segmented deformable secondary, with segment geometry similar to that of the primary. These segments, which build on the technology developed at the MMT and LBT telescopes, will be highly agile in tilt and piston motion as well as in deformation. Each one will be used to servo out rigid body motions of the primary segment to which it conjugates, as well as atmospheric distortion. Realization of the new wavefront sensing strategy that requires a constellation of laser beacons has been a concern. There is still little experience with even single beacon systems because the sodium-tuned lasers have proven so challenging. For this reason Arizona is pioneering laser tomography with simple YAG lasers at the existing MMT telescope, and has recently shown for the first time tomographic reconstruction of a stellar wavefront from a multi beacon constellation [3]. The GMT design incorporates the projection and sensing strategies already proven at the MMT, carried over to the sodium lasers needed for the much larger aperture. 611

3 2. OVERALL DESIGN OF THE GMT Fig.1a shows a general view of the telescope. The optical design is an aplanatic Gregorian, with an 18 m focal length primary in a 25 m envelope and a 3.3 m secondary that forms an f/8.4 focus 5.5 m behind the primary vertex. The seven primary mirror segments are circular, each one 8.4 m in diameter, with 30 cm gaps between. The secondary mirror is held 20 m above the primary vertex by a rigid structure designed to avoid any obscuration of the six outer segments (Fig.1b) and to have low wind cross section. The supporting structure consists of three tripods surmounted by a hexapod. The first of the ring of six identical off-axis primary segments has been cast as a honeycomb sandwich of borosilicate glass, like the primary mirrors of the 3.5 m Starfire Optical Range, ARC and WIYN telescopes and the larger MMT, Magellan and LBT primaries. In use, the honeycombed substrates are ventilated so as to match the ambient air temperature. This largely eliminates mirror seeing, as shown by the seeing limited images of less than 0.5 arcsec FWHM that are common at the Magellan telescopes, with occasional entire nights of 1/3 arcsec. The presumptive site for the GMT is the slightly higher Las Campanas peak near Magellan. We are confident that the highly aspheric, off-axis figure of the six identical outer segments of the very fast f/0.7 surface can be produced by the stressed lap polishing method. This method has been proven in the manufacture of the two f/1.14 LBT primary mirrors, with the largest aspheric departure of any mirror yet made. The GMT segments require no larger bending stress levels on the lap [4]. A 1/5 scale prototype is being figured now using inprocess metrology similar to the that being implemented for the full-sized segment [5].. In the diffraction limit, the entrance pupil with seven circles results in a beam profile with six-fold symmetry, as shown in Fig. 2. The central peak has a FWHM the same as for a clear circular aperture of 24.4 m, and contains 67% of the incoming light. This is similar to a circular aperture with 10% area obscuration, for example the HST places 72% in the central peak. A completely unobscured circular aperture core has 84%. Fig. 2. Diffraction limit PSF for a 24.4 m filled disc (left) and the GMT (right). Both have the same FWHM, for example arcsec at 1.65 µm. The GMT provides for interchangeable large instruments (up to about 250 cubic meters) to be located at the direct Gregorian focus where the corrected field is 20 arcminutes diameter. Multiobject spectrographs operating at the seeing limit and with AO ground layer correction up to a maximum field of 8 arcminutes will be at this location. Most AO instruments require smaller field and will be located on a 9 m diameter platform below the primary mirror, mounted so as to rotate about the telescope axis to compensate for field rotation. Articulated tertiary mirrors will allow these to be brought into play quickly when desired, as described in section 7 below. 3. SEGMENTED DEFORMABLE SECONDARY MIRROR The Gregorian secondary mirror takes the form of seven circular segments, each conjugated to the corresponding primary segment. The secondary mirror acts as the main, deformable correcting element for the GMT adaptive optics system. Each segment will be deformable with 672 actuators, spaced at 30 cm relative to the incoming wavefront. These parameters are the same as for the LBT deformable secondaries, which along with the MMT deformable secondary serve as prototypes [6,7]. The GMT s facesheets will be a bit larger and thicker, 1.1 m diameter and 2.4 mm thick compared to 0.91 m and 1.6 mm respectively for the LBT. Each of the seven facesheets is held in position with reference to its own underlying rigid reference body. Annular capacitive sensors around each actuator are used to measure the gap of ~ 100 µm with a precision of a few nm at a local servo update rate of 90 khz. This is fast enough that damping of high frequency vibrations can be accomplished electronically. The larger gap allows a larger stroke than for the MMT adaptive secondary prototype, 612

4 Fig. 3. Detail of top end. The main hexapod tower terminates in a 3-point star structure with the sodium constellation projection lens at its center. The adaptive secondary is held below by an adjustable hexapod (shown in red) to compensate for telescope flexure. The seven segments of the secondary are joined by a hexagonal truss with extended points on each corner. Fig 4. View of the segmented secondary from an instrument. The support structure and laser projector optics are arranged so as to be largely hidden behind the seven segments. Rayleigh scattering shown from the laser beams will appear only in the central segment of the pupil. where high frequency damping is by the viscosity of air in a narrower (40 µm) gap [6]. The secondary stroke will be large enough to bring the entire aperture into phase by correcting for dynamic tip, tilt and piston terms as well as higher order terms. Additional control capability for the secondary segments will be incorporated in the small hexapods that link the reference bodies of Zerodur to the single 3-m supporting carbon fiber truss. These hexapods will have rigid legs of piezo or magnetostrictive material, with adjustment of ~ 100 µm relative to the truss. The truss itself is connected to the main secondary tower via a single, large hexapod with travel of several mm, enough to correct for slow gravitational and thermal displacements of the secondary tower relative to the axis of the primary mirror (Fig. 3). One of the reasons for choosing the Gregorian configuration for the GMT (and previously at the LBT) is to enable closed loop AO system tests to be made at any time. This will be done with a star simulator at the top end which can be moved by remote control into position at the prime focus. This will be especially useful for daytime engineering of the closed loop AO system, a feature that is sorely missed at the MMT prototype, which is Cassegrain. It will be used also to align the secondary segments to be parts of a continuous surface, and to develop open loop secondary flexure corrections. The prime focus source will include a dynamic aberrator to reproduce images corresponding to different seeing conditions. 4. ACTIVE CONTROL STRATEGY Underlying the adaptive optics control system will be a slower, active system, designed to correct large flexural and thermal distortions and hold the combined image for negligible degradation of the best seeing limited images. For the individual monolithic glass primary segments, flotation systems to minimize gravitational bending and ventilation to minimize thermal distortion are already developed and proven in existing honeycomb primary mirrors such as Magellan s. The overall mirror figure of the GMT, though, is set by the 25 m steel structure to which the primary segments are attached by hexapods. This structure is not supported in flotation and is not thermally stabilized. As the telescope is moved in elevation and is subjected to wind loads, differential motions of the primary segments are projected to cause about 1 arcminute separation of the images from the 7 segments. Thermal distortion 613

5 will also contribute to the misaligment. To preserve in the combined images the best seeing of 0.25 arcsec, stabilization to better than 0.1 arcsec is needed. The GMT is unique among segmented mirror telescopes in that misalignment of the primary segments can be corrected by compensating motions of the conjugated secondary segments. Such correction will be especially useful to trim dynamic primary misalignments as might be caused by wind gusts. In principle, even the full range of misalignment can be corrected entirely by moving the secondary segments, with no significant loss in image quality over the 20 arcmin field [8]. Our preferred strategy is to control the primary segment positions to better than 10 microns, roughly a factor of 15 smaller than the gravity deflections, by frequent adjustments to the hexapods that support each segment in steps of less than a micron. Residual errors will be corrected by the secondary segments. The information needed to make these corrections will be obtained from field stars over the 20 arcminute field, viewed by Shack Hartmann sensors across the full pupil. In addition, edge sensors will measure the relative displacement of adjacent primary segments with high frequency response, as for the segmented Keck telescopes. To bridge the 30 cm gap between segments, the GMT s sensors will be optical. Because the active system will remain in operation during adaptive optics use, any spatial and temporal discontinuities that are introduced by corrections must be minimized. The primary hexapod actuators will be designed along the lines already developed for the LBT to operate smoothly, and in any case the dynamic response of the segment will be characterized and monitored during operation with the edge motion sensors, so that compensating motions of the deformable secondary segments can be made automatically. 5. ADAPTIVE OPTICS CONTROL SYSTEM The primary wavefront sensing system will use a constellation of 5 sodium laser guide stars, each viewed across the full pupil with a fast Shack Hartmann sensor. Field stars will be sensed in the optical and near infrared to provide tip/tilt and piston components averaged across each wavefront segment. Correction will be applied with the adaptive secondary, thus ensuring the highest throughput and lowest thermal background. The two basic modes of operation implemented with this system will be Laser Tomography (LTAO) and Ground Layer correction (GLAO). The third basic mode will be for high contrast imaging, in which additional use will be made of natural star sensors in both the pupil and focal planes. Fig. 5. Tip/tilt and piston sensor. The left hand column shows three different pupil configurations used to form images. First an annular beamsplitter in the pupil redirects half the light from the ring of six outer segments to form the bottom image in the second column. The remaining light is split again to form the two upper pupils and images. The MTFs of the three images so formed are shown for the cases of piston error in the 9-oclock segment (third column) and in the center segment (fourth column). For diffraction limited imaging via LTAO, the beacon constellation will be set at 1.5 arcminutes diameter, so as to sample turbulence at all heights across the full aperture. The wavefront for a star on-axis will be computed by tomography from the wavefront data. Such tomographic reconstruction from multiple beacons is being pioneered by the Center for Astronomical Adaptive Optics at the MMT, with the latest results given elsewhere in these proceedings 3. For the GMT, wavefront reconstruction will be extended to bridge the discontinuities across the segment gaps. This will be accomplished (along with tip/tilt sensing) by analysis of infrared focal plane images of faint field stars, the same as used for tip/tilt measurement. The images will be recorded simultaneously through different blocked pupil configurations, as shown and explained in Fig. 5. The field of view for correction to the diffraction limit when correction is made at a single deformable mirror conjugated to the ground is 1 arcminute, depending on wavelength, as for current AO systems. This depends on wavelength and the vertical distribution of turbulence, but not on telescope aperture. Increased field at the diffraction limit should be possible by use of additional optics with deformable mirrors conjugated to higher levels of turbulence (MCAO). Such correction, still a few years from being proven on any large telescope, is envisaged as a later addition to the GMT. 614

6 GLAO provides a way to obtain improved image quality over a field of view of several arcminutes by correcting with the deformable secondary alone the turbulence near the ground. Because the GMT s secondary is conjugated to 150 m above the ground, the correction will hold over a large field of view. To measure the ground layer aberration, the diameters of both the laser constellation and the wavefront sensing detectors must be spread out over the field to be corrected, which could be as large as ~ 5 arcminutes, depending on the thickness of the ground layer. In the first measurement of ground layer wavefront distortion made with the MMT multi-beacon system, the ground layer was at a median height of 400 m. Under these conditions, the data show useful correction extending over several arcminutes [9]. The ground layer correction signal is derived by averaging the wavefront taken over all the beacons, or better from the full tomographic solution. Provisions at the GMT for projecting and sensing the beacon constellation over the range of diameters needed for LTAO and GLAO are described in the next section. 6. SYSTEM FOR VARIABLE DIAMETER LASER CONSTELLATIONS The overall layout of the projection system is shown in Fig. 6. The 5 sodium lasers reside in a thermally controlled enclosure on the azimuth platform. The beams are directed by two mirrors on the elevation axis to a projection lens above the secondary mirror. CW lasers tuned to the sodium resonance line of the high power and reliability required for AO have only now been perfected [10]. The pulsed formats preferred for control of spot elongation with very large aperture are under development, but today we cannot say how much room they will require. As a guess, we take a size appropriate for today s CW systems, assuming vertical lasers with a 0.4 x 0.9 m footprint, and power and cooling in a rack with a 0.9 x 0.9 m footprint. Fig. 7 shows the layout of lasers and the beam-directing mirrors in the Fig. 6. View showing the laser optics. Five sodium laser beacons are housed in the room on the az-platform (lower left). The beams are directed to a projector above the secondary via two flat mirrors on the elevation axis. The platform-mounted AO instruments are visible directly below the primary mirror assembly. Fig. 7. Plan of lower floor of laser room. The 5 lasers are arranged vertically along one wall, the power and cooling racks opposite. The beams emerge from the lasers pointing up. They are redirected by 90 degree fold mirrors in the ceiling as shown so as to form a pentagon constellation of 5 beams traveling horizontally to the left. The mirrors are on adjustable linear stages so the size of the pentagon size can be controlled without losing the pure linear polarization of the beams, or the apparent beam waist distance, which is the same for all five. lower floor. Also shown in the figure is the system of four 90 degree reflections per laser than brings the beams parallel in a pentagon of variable diameter from 60 to 300 mm, corresponding to 1 5 arcminutes on the sky. These beams are directed through the upper story enclosure to the elevation axis mirrors. The enclosure houses diagnostic 615

7 Fig. 8. Beam conditioning optics in the upper floor. The five beams are directed by M1 toward the elevation axis. Their polarization is first conditioned by passage through slightly tilted uniaxial crystals, so as the make the projected beams circularly polarized for strongest return. Rotation of the constellation on the sky is then accomplished by a K mirror. A beamsplitter then takes about 0.1% of the beams for diagnostics, and the beam emerges through a thermally insulating window. optics and the means to rotate the constellation and condition the beam polarization shown in Fig. 8. Two mirrors on the elevation axis direct the beams to the 0.5 m telephoto projection lens above the secondary, which reimages the diffraction limited beam waists formed in the lasers onto the sodium layer at a range of km, depending on telescope elevation. The lens has a back focal distance of 40 m and a focal length of 150 m. This sets the magnification that projects a 5 arcminute constellation for 220 mm separation of the beams at the laser house. The five beams are directed slightly inward so their chief rays meet at a common pupil with ~10 cm waist formed at the telephoto lens. The mirror that receives the beams above the secondary and directs them to the central axis is located at this pupil. It will be equipped with fast tip/tilt actuators to correct for common seeing or vibration-induced jitter in the projected constellation. Our overall sensing configuration for the GMT is designed so that one sensor suite will serve for all AO focal stations. This avoids having to duplicate the complex system which requires each of the five laser wavefront sensors to detect patterns with ~ 5000 Shack Hartmann spots at khz rates and to be articulated so as to accommodate different constellation diameters and ranges. For the preferred pulsed laser beacons the sensors must also include gating and refocusing to minimize spot elongation, as implemented at the MMT tomography system [9]. This sensor system will be fed by a dichroic tertiary mirror on a calcium fluoride substrate reflecting optical wavelengths from the central 8 arcminute field. It will be located on the rotating platform above the direct Gregorian focus, as shown in Figs. 9 & AO MODES AND INSTRUMENT LOCATIONS AO instruments for the near IR can be deployed at the direct Gregorian focus, below the dichroic that feeds the optical wavefront sensors, or at stations fed by additional tertiary dichroics above the optical wavefront sensor. The optical diagram is shown in Fig. 9. The design is such that the quasi-static active optics sensors are always in play, operated over an unvignetted 10 to 20 arcminutes annular field at the direct Gregorian focus. The four AO modes to be implemented initially are listed in table 1. For GLAO wavefront measurements only the Low dichroic will be in play, feeding both the five beacon sensors and optical tip/tilt sensors using faint natural guide stars over the 8 arcminute field. The science instrument will be at the direct Gregorian focus (mode 1). In mode 2 diffraction limited instruments in the near IR will be located on the upper platform and fed by a mid-level, additional dichroic tertiary mirror, this one reflecting the infrared and transmitting the optical. In this way the sodium beacons Table 1 observing modes Mode Low Middle Upper Notes dichroic dichroic dichroic No AO retracted out out Sensor unit retracts to pass full 20 field 1 GLAO in out out 8 NIR field passed by CaF 2 dichroic 2 LTAO NIR in in out 4 NIR field reflected to center instrument in Fig. 8 3 LTAO MIR in in in 4 MIR field reflected into right hand instrument 4 High contrast AO in in out Optical NGS sensing fed by low dichroic, additional sensing in NIR unit 616

8 Fig. 9. Layout of dichroic tertiary mirrors to feed the optical sensors (lowest) and infrared sensors and science instruments. Fig. 10. Sensor suite and instruments on the upper rotating platform. The sensor suite with 5 Shack Hartmann laser beacon sensors is fed by the lowest dichroic and housed in the left hand unit. will be detected by the same sensors, via the low dichroic as before. Tip/tilt and phase sensors in the near IR instrument dewar will be fed by small articulated pick-off mirrors close to the NIR science focal plane. Several instruments may share the same pick-off system. In mode 3, thermal infrared instruments will be fed by another, high tertiary dichroic which will transmit NIR and optical light to the same laser beacon and near IR tip/tilt sensors below. In all the above modes individual science instruments will likely include at least one internal tip/tilt sensor to compensate for flexure. High contrast near and mid-infrared imagers (mode 4) will be located according to their wavelength and need for the system sensors. Faint primary stars such as white and brown dwarfs may benefit from the telescope s beacon wavefront sensing system. Aberrations of bright primary stars may be sensed entirely by their own light, as we indicate in the next section. 8. HIGH CONTRAST AO FOR EXOPLANET IMAGING The GMT project has set a high priority on high-contrast imaging of exoplanets. The telescope s adaptive secondary will again be used for the AO correction. Three key elements make up a high contrast system. First, an optical wavefront sensor operating at high frame rate and resolution is used to control the secondary to recover a high Strehl image. For bright stars the Strehl will be high enough that the pupil diffraction pattern will become the main halo contribution, and this must be suppressed by a second element, to reduce photon background noise. The third element is an infrared focal plane sensor as described below, to derive improved suppression from interferometric measurements of the residual atmospheric speckles. For the second element, we do not plan to use a conventional Lyot coronagraph, which is especially inefficient for the segmented pupil of the GMT, but the new method of Codona and Angel [11]. This avoids altogether the use of a Lyot stop and the associated losses in throughput and resolution. Suppression of the diffraction pattern is accomplished simply by an additional modulation of the telescope s deformable secondary, with no need for additional optics. The deformation is chosen to diffract a detailed opposite of the pupil diffraction halo, with the same amplitude but opposite phase. This can be done for up to half the field (i.e. a 180 degree wedge). Fig. 10 shows the nulled GMT PSF modeled for 10% bandwidth with modulation of the 4600 actuator GMT deformable secondary chosen for search mode, with suppression over 180, Fig. 11a shows the PSF in the absence of atmospheric or other aberrations. The suppression causes little reduction of resolution, and a loss of only 33% of 617

9 light from the central core, much more benign that coronagraphic halo suppression. The dark semicircle is the largest diameter (1 arcsec in the H band) correctible with the adaptive secondary alone. The outer limit of suppression set by the limit to high frequency components that can be controlled because of the finite number of actuators, and is proportional to wavelength. Fig. 11. PSF s for high contrast imaging, obtained by modifying the figure of the 4600 actuator deformable secondary, modeled for search mode in H band with 10% bandwidth. a) Left diffraction limit for no atmospheric distortion. (b) Right With the addition of residual aberration of the AO system represented by a 2 cm offset between the wavefront and the correction figure. Seeing given by r 0 =15 cm at 0.5 µm wavelength. The ideal PSF of Fig. 11a will be degraded by residual wavefront errors left by the AO system. The most significant of these is likely to be servo lag. This may be modeled by supposing correction applied to the deformable secondary represents the wavefront as it was a short time earlier. An expression for the resulting PSF is given by Angel [12] for a distorting Kolmogorov layer with Taylor drift. A numerical realization by Codona of an instantaneous PSF is shown in Fig. 11b. Here the wavefront coherence length is taken as r 0 =15 cm at λ= 0.5 µm : i.e. not especially good seeing. The translation (drift speed times servo lag) was taken to be 2 cm, with the drift direction to the right. This translation is appropriate for a wind speed of 20 m/sec and 1 msec servo delay. This is realistic for the adaptive secondary with its response time of 0.5 msec. Fig. 12 (left) shows the instantaneous radial PSF averaged over the 90 sector about the drift direction, where the halo is strongest. It ranges from 2x10-5 at 0.1 arcsec to 3x10-6 at 0.4 arcsec. A different optimization of the adaptive secondary modulation for follow-up mode was made with suppression over 60 only, but with closer inner working angle. The core energy loss in this case is only 25%, and the radial PSFs averaged over 60 are shown in fig 11b. We gain access very close to the star, with a residual halo level averaging 8x10-5 at 0.05 arcsec (3.6 λ/d eff ) and 1.3x10-5 at 0.1 arcsec. The limiting contrast for long exposures will depend on how the stellar speckle background averages out over time. Current high contrast images with AO systems show persistent speckles. We plan to remove both these and the changing atmospheric speckles by use of the third key element, which is an additional servo based on a focal plane interferometer to measure at high speed the complex amplitude of the speckles at the science imaging wavelength [11,12]. A second servo loop will then be applied to modulate the secondary (based on a Fourier transform) to null out the speckles. A predictive tracking algorithm that produces statistically independent speckle patterns every millisecond should be possible. Assuming that the signal to noise for a planet at the average speckle level is unity in each millisecond image, the 5 sigma signal to noise ratio after 10 7 milliseconds (3 hours) will be 630 times lower than the instantaneous PSF shown, i.e. 3.2x10-8 at 0.1 arcsec to 5x10-9 at 0.4 arcsec in search mode and 1.3x10-7 at 0.05 arcsec to 2x10-8 at 0.1 arcsec in follow-up mode. Such sensitivity brings within reach known giant exoplanets seen in reflected starlight. A technological advance is needed for such high contrast imaging, namely the fast, nearinfrared focal plane detector with very low read noise that serves as both the speckle and science sensor. The GMT partners plan to develop such detectors and test the suppression method with the MMT and LBT AO systems. 618

10 Fig. 12. Instantaneous average radial profiles. Left - For the PSFs of figures 10a and 10b, averaged over the brightest 90 wedge (1:30 to 4:30-o-clock). Right The same but with the adaptive secondary suppression reoptimized for close-in detection, averaged over a 60 wedge. Higher sensitivity may be possible for the brightest candidate stars around 5 th magnitude. Reduced photon noise in the wavefront sensor may allow for better correction by running the servo at cycle times faster than 2 khz, the deformable secondary limit. For these, an auxiliary instrument will be used at the GMT, including a deformable mirror with faster response and but modest stroke. 9. CONCLUSION A giant segmented mirror telescope was recommended 5 years ago by the decadal Astronomy Survey Committee of the National Academy of Sciences. The GMT design we have developed and outlined above provides a system with integrated adaptive optics, capable of fully exploiting the potential of very large aperture to image at the diffraction limit. Its primary mirror exploits the technology and production facilities developed for large borosilicate honeycomb mirrors and proven at the MMT, Magellan and LBT telescopes. Its secondary mirror is deformable, built of segments similar to the proven prototype at the MMT and the new LBT secondaries. Conjugation of the primary and secondary segments provides a robust solution to wind excitation of the large structure. The Gregorian secondary is fully sky baffled, and should give excellent performance with adaptive correction in the thermal infrared, as demonstrated with the first deformable secondary at the MMT. The GMT s multi-beacon laser system has been prototyped at the MMT and been shown capable of reconstructing ground layer and tomographic wavefronts. At the best seeing sites in Chile where the GMT will be located, the lower wavelength limit for diffraction limited imaging of faint sources will be around 1 µm, set by photon noise in the laser wavefront sensing system and the speed and strength of atmospheric turbulence. The one location on the planet where atmospheric conditions are markedly better is the high Antarctic plateau. Here the coherence length (Fried s length) and coherence time at 0.5 µm wavelength appear to be similar to those in Chile at 1 µm [13]. It follows that a copy of the GMT with the same AO system would achieve diffraction limited imaging at optical wavelengths, with ten times the resolution of the Hubble telescope. The clean thermal design of the GMT would also ensure that the unique thermal advantages of the Antarctic could be also fully exploited [14]. By building a close copy of a telescope already tested at a temperate site, the difficulties of construction and engineering in the short Antarctic season would be minimized. The GMT project has no plans at present for such a copy, but we can dream. 619

11 ACKNOWLEDGEMENTS This work is supported by the AFOSR under grant F and the NSF under grant AST We are grateful to Phil Hinz and Steve Shectman for their help in developing the GMT AO concept, and to Matt Rademacher, Miguel Snyder and Nicole Putnam for their assistance in mechanical and optical design. REFERENCES 1. Johns, M., Angel, R, Shectman, S., Bernstein, R., Fabricant, D., McCarthy, P., Phillips, M., Status of the Giant Magellan Telescope (GMT) Project, Proc SPIE, Vol. 5489, , Gunnels, S., Davison, W., and Cuerden, B., The Giant Magellan Telescope (GMT) Structure, Proc SPIE, Vol. 5495, , Lloyd-Hart et al., Tests at the MMT of Multi-laser Guide Star Wavefront Sensing for Advanced Adaptive Optics, these proceedings, Martin, H. M, Allen, R. G., Burge, J. H., Dettmann, L. R., Ketelsen, D. A., Miller, S. M. and Sasian, J. M., Fabrication of Mirrors for the Magellan Telescopes and the Large Binocular Telescope, Proc SPIE, Vol 4837, , Martin, H. M., Burge, J. H., Cuerden, B., Miller, S. M. Smith, B., Zhao, C., Manufacture of 8.4 m Off-axis Segments: a 1:5 scale Demonstration, Proc SPIE, Vol. 5494, 2004 and Martin, H. M., Angel, J. R. P., Burge, J. H., Miller, S. M., Sasian, J. M. and Strittmatter, P. A., Optics for the 20/20 telescope, Proc SPIE, Vol. 4840, , Wildi, F., Brusa, G., Lloyd-Hart, M., Close, L. and Riccardi, A., First Light of the 6.5-m MMT Adaptive Optics System, Proc SPIE, Vol. 5169, 17-25, Riccardi, A., Brusa, G., Salinari, P., Busoni, S., Lardieri, O., Ranfagni, P., Gallieni, D., Biasi, R., Andrighettoni, M., Miller, S., Mantegazza, P., Adaptive secondary mirrors for the Large Binocular Telescope, Proc SPIE, Vol. 5169, , Johns, M., GMT Optical Alignment Calculations, GMT Document No. 1297, Lloyd Hart, M., Baranec, C., Milton, M., Stalcup, T., Snyder, M., Putnam, N. and Angel, R., First Tests of Wavefront Sensing with a Constellation of Laser Beacons, ApJ, accepted for publication, Denman, C., Recent Results Using the 50W Sodium Guidestar Pump Source at the Starfire Optical Range, these proceedings, Codona, J., and Angel, R., Imaging Extrasolar Planets by Stellar Halo Suppression in Separately Corrected Color Bands, ApJ, Vol. 604, L117-L120, Angel, R., Imaging Extrasolar Planets from the Ground, in Scientific Frontiers in Research on Extrasolar Planets, ASP Conference Series, Vol. 294, , Lawrence, J.S., Ashley, M.C.B., Tokovinin, A. & Travouillon, T. Exceptional astronomical seeing conditions above Dome C in Antarctica, Nature Vol 431, , Angel, R., Lawrence, J. and Storey, J., Concept for a Second Giant Magellan Telescope (GMT) in Antarctica, Proc SPIE, Vol. 5382, 76-84,