 |
 |
A. Tokovinin, H. Schwarz
Draft 1.6. August 7, 2001
Most modern large telescopes are now being equipped with Adaptive Optics (AO) systems. Next door to SOAR, Gemini is planning two systems, a near-infrared coronographic imager (NICI) and an ambitious multi-conjugate AO (MCAO) with laser guide stars (LGSs). One of the 8m VLT telescopes will have a working AO (NAOS) in 2001 (with more AO systems to follow) and for a while will have exclusive capabilities on the southern sky. How does SOAR fit into this picture of modern ground-based astronomy with 8m class telescopes using sophisticated AO systems?
It would make little sense to compete with Gemini and VLT in the near infrared. If both 4m and 8m telescopes are equipped with AO systems, diffraction-limited resolution at 8m would be 2 times higher, and sensitivity 4 times better than at SOAR. As S. Ridgway noted, with AO ``the rich get richer and the poor get poorer'' [2]. However, there is no matching effort in the visible-range AO, where SOAR may have its unique scientific niche. In this document we will therefore consider optical AO possibilities for SOAR.
Over the next few years both SOAR and Gemini S will be commissioned and put into science operations. The coherent plan for Tololo and Pachón will optimize the 4m Blanco for wide field imaging & spectroscopy, and SOAR for high resolution imaging & spectroscopy, leaving 8m class science to Gemini.
Gemini will clearly do it's best, high redshift science in the near and mid IR over a small to medium FoV of 1-5'. SOAR was conceived with built in tip-tilt correction, and will do z=zero science at the highest possible spatial resolution over a medium FoV of up to 5' or so, while the Blanco will use a FoV of up to 40' to do imaging and multiple object spectroscopy at medium image quality.
SOAR equipped with an AO system working in the visible range will form a powerful complementary part of the SOAR, Blanco, Gemini system
Although the technology of adaptive optics has entered into maturity, turbulence compensation still remains difficult and technically very challenging. The two successful astronomical AO systems (Adonis at ESO and PUEO at CFHT) need relatively bright natural guide stars (NGSs), severely limiting the choice of astrophysical targets. Laser guide stars (LGSs) alleviate this problem only partially, because for long-exposure imaging there is still a need to compensate tip-tilt with NGSs. Even with a very bright LGS, it would therefore not be possible to achieve diffraction-limited imaging in the visible range covering the whole sky. To date, no significant astronomical results have been obtained with LGSs.
Given the high investments needed for the implementation of astronomical AO systems, and relatively modest scientific returns to date, some scepticism has developed in the community concerning the usefulness of AO. If SOAR is to be equipped with AO beyond tip-tilt, a very careful evaluation of it's expected performance and scientific case is mandatory, as well as a cost benefit analysis of the various possible systems.
The challenge of SOAR AO project lies not in the blind cloning of existing AO systems, but in creating a really useful and efficient instrument that can be scheduled a significant fraction of time and can serve a wide range of astronomical programs. Being a second-generation AO system, it must take full advantage of all previous experience and improved and-or cheaper technology. The cost must be kept to a reasonable proportion to the overall cost of the SOAR project; SOAR is a 4m class telescope, not a Keck or Gemini.
A 4m class telescope will soon be considered ``small'', and, like other small telescopes, will have most of its scientific impact through large surveys, synoptic and follow-up observations, etc. A highly efficient, cost effective operating model with a fixed complement of rapidly switchable instruments, is already incorporated into the current SOAR design and will have to evolve into a full queue scheduled service observing model in the near future. Queue scheduling is almost imperative in order to use efficiently the periods of best seeing, as planned for SOAR. This efficient use of the rare periods of good seeing is accentuated even more with AO, because its performance critically depends on seeing. Hence, the SOAR AO system must be suited to queued observing. This is a challenging task: in case of the VLT, for example, the consistent implementation of this principle has led to the costly decision to build separate AO systems for each instrument. Even if SOAR were equipped with an adaptive secondary, it would not be possible to ``serve'' the whole complement of SOAR instruments with a single AO system, because separate wave front sensors would be needed at each focal station (the same holds for the tip-tilt tertiary mirror). The SOAR AO concept must specify which instruments will be used with AO and how they will be switched.
An AO project for SOAR, scientifically valuable in itself, must also be evaluated in the context of the next generation of Extremely Large Telescopes (ELTs) with primary mirrors in the range 30-100m. If NOAO is to play a major role in this development, it is vital to form an engineering team capable of developing AO systems and gain practical experience in AO as soon as possible since such experience is lacking within NOAO at present. Some of the approaches suggested below for SOAR may be scalable to ELTs.
In the following chapters, estimates of the expected AO performance
are given, based on the modal simulation code [6] and current knowledge about the
Pachón seeing ([8] and Table 1). These estimates should be taken as
preliminary, they serve to identify promising system configurations
and parameters rather than to give ultimate and precise numbers.
| Probability | 25% (good) | 50% (median) | 75% |
| Fried parameter at 0.5 micron, cm | 20.0 | 15.0 | 10.7 |
| Standard FWHM at 0.5 micron, arcsec | 0.50 | 0.67 | 0.94 |
| Finite-outer-scale FWHM at 0.5 micron, arcsec | 0.40 | 0.55 | 0.79 |
| Finite-outer-scale FWHM at 2.2 micron, arcsec | 0.24 | 0.33 | 0.49 |
The SOAR design already incorporates a first order turbulence compensation by providing a fast motion of the flat tertiary mirror. Initial estimates of the improvement in image quality achieved by tip-tilt (tt) compensation were based on the preliminary seeing data for Cerro Pachón and on the standard turbulence theory with infinite outer scale. Now these estimates should be revised in view of the detailed characterization of the Pachón atmosphere [8]. A summary of this report is provided separately. Briefly, the seeing is slightly worse than originally believed but still comparable to the best sites in the world, Table 1. The knowledge of the finite turbulence outer scale permits a better prediction of the Point Spread Function (PSF) width at different wavelengths (it is smaller that in standard theory) and at the same time reduces the expected gains from tt compensation. The knowledge of the vertical turbulence profiles leads to a revision of the ``isokinetic patch'' concept: the degradation of PSF with increasing distance from the tt guide star (GS) can now be quantified.
|
|
|
The reader may consult detailed simulation results. In Fig. 1 two representative cases are illustrated. The uncompensated PSFs are plotted, followed by the traces of PSFs at the position of the GS and at increasing distances from the GS within the field of view (FoV). The left panel shows the results for median seeing in the visible range (all r0 are given for a nominal wavelength of 0.5 micron), where the gain from tt is only marginal. In the right panel, the case of strong gain (K-band imaging, good seeing) is illustrated. Evidently, the gain is maximum at the position of GS and decreases rapidly going away from the GS. There is considerable PSF non-uniformity over the FoV, and already at 1' radius the degradation is significant.
Uniform tt compensation over the whole FoV but with reduced gain can be achieved by using 3 GSs instead of one. If these GSs are located around the periphery of the FoV and their signals are averaged, only the tt produced by the lowest turbulent layers is compensated. No SOAR instrument is designed to use more than one GS for tt compensation, however. The consequences of "one-GS tt" in terms of PSF uniformity over FoV must be carefully evaluated.
These simulations do not take into account telescope wind shake that will also be compensated by tt, increasing the actual resolution gain with respect to the idealized model. As yet, the amount of shake is not known, and it may turn out that SOAR structure is stiff enough to make it insignificant. If shake is a problem, telescope vibrations in the frequency band 1-30 Hz can be measured in real time by an inertial sensor and fed to the tertiary. The sensor with adequate characteristics has been prototyped [6], and it's implementation is inexpensive.
Thus, an alternative strategy for tt compensation would be the use of a seismometer to remove vibrations (if necessary) and slow (less than 1 Hz) guiding on a star, which can be 2 magnitudes fainter than for fast atmospheric tt compensation. Uniform PSF over the whole SOAR FoV will be assured, because slow atmospheric tt components are highly correlated. Whenever the highest resolution in a reduced FoV is needed, a fast tt should be used, as originally planned. The largest resolution gain is expected in the IR, where the light from GS is picked off by a dichroic mirror, permitting to select a GS within the scientific FoV or to guide on the object itself. Under median seeing, the on-axis Strehl ratio in the K band is 0.08 without tt and 0.21 with tt. Under good seeing these numbers increase to 0.15 and 0.37 respectively. Building an IR imager with diffraction limited sampling is fully justified even without AO, although a modest additional compensation beyond tt will provide a significant improvement of Strehl ratio.
A low order AO system is the next logical step beyond simple tt. With only 19 actuators, the PUEO AO at the 3.6m CFHT telescope has achieved an impressive performance ([5], Ch. 9). The size of individual sub-apertures is about 1 m. Such an AO system can be viewed as a phased array of 1 m telescopes. A useful degree of compensation can be achieved with NGSs of 15th magnitude.
A low order AO provides diffraction limited images in the JHK bands. It also improves seeing in the visible range. This effect can be understood by noting that image size in a 1m telescope with fast tt guiding is reduced. The maximum effect of tt is achieved at D/r0~4 which corresponds to sub-apertures of 60 cm for median Pachón seeing at 0.5 micron. In Fig. 2 the shrinking of the PSF halo is quantified in a system independent manner. Maximum gain in Full Width at Half Maximum (FWHM) is a factor of 2 and corresponds to a Strehl ratio of only few percent, but the gain persists even at vanishingly small Strehl ratios.
 |
Figure 2: The gain in FWHM (solid line) and in half-energy diameter FWHE (dashed line) with respect to "standard" (infinite-scale) seeing limit for the halo of partially compensated PSF, as a function of d/r0, the ratio of sub-aperture diameter to Fried parameter. The corresponding Strehl ratio is traced in dotted line. |
Thus, a low order AO system offers a useful performance gain in the visible range and can operate on a 15th magnitude NGS. Still, the problem of PSF variation over the FoV remains. The diameter of the compensated FoV is twice the isoplanatic patch in the IR, and it may reach 30" and more, but the compensation quality will degrade towards the edge of FoV. If a GS lies at some distance from the scientific target (as would be often the case), the performance is significantly reduced, which effectively limits the sky coverage.
A solution to this problem consists in using several guide stars and several deformable mirrors. This approach, called MCAO, was chosen for the Gemini south AO system. It is not yet demonstrated on the sky and is expensive. MCAO works well when there is a significant overlap of the GS beams in the upper atmosphere. For Gemini, the achievable FoV diameter is about 1'. In such a small field, there will be not enough bright NGSs, forcing the use of LGSs in order to achieve a decent sky coverage. For SOAR, with half the mirror diameter, the same beam overlap corresponds to half the FoV (30"), which is restrictive not only for sky coverage but for the potential benefits of such a system as well. Gemini style MCAO is not an interesting perspective for SOAR.
Instead of trying to compensate the whole turbulent volume, we could compensate the ground layer turbulence only [3]. To do this, a number of widely spaced GSs must be selected to sense the wave fronts. Averaging these signals, we isolate the ground layer contribution which is the same for all GSs, and ignore the contribution of higher layers. The achievable degree of compensation will be less, but it will be uniform over the whole FoV spanned by the GSs. This is the simplest case of turbulence tomography.
In a FoV of 4-6' diameter three NGSs of 15th magnitude can be found with high probability. In [1] the sky coverage is estimated to be 10% at the Galactic pole and 100% at lower Galactic latitudes. There will be no problem to find even brighter and closer NGSs in the Galactic plane and in the Magellanic clouds. By selecting brighter and closer NGSs, we reduce the compensated FoV but increase the resolution gain. Thus, there is a trade-off between FoV size and the degree of compensation. A combination of parameters (three 15th magnitude GSs, 4-6' diameter FoV, 1 m sub-apertures) provides a useful zero-order approximation for a realistic system, with a possibility of further optimization around these ``canonical'' values depending on the region of the sky under study and the scientific goals (resolution vs. FoV).
 |
Figure 3: PSFs at 0.5 micron and median seeing for uncompensated imaging (left), on-axis, and at increasing distance from the guide star for an AO system with a Rayleigh LGS at 10 km altitude. |
Rayleigh LGSs are formed at low altitudes (10-25 km) and are known to be insensitive to turbulence at high altitudes due to the cone effect. This is precisely what is needed for ground layer compensation! By lighting an LGS at an altitude of 10km, we can achieve a uniformly compensated FoV of 2-3' diameter (Fig. 3). For Rayleigh LGSs, cheap and reliable lasers radiating in the UV are used. Low altitude increases the return flux and relaxes requirements for laser power, thereby eliminating the risk to aircraft and satellites with it's associated legal and operational complexities. Such a Rayleigh LGS AO system will be cheaper and more practical than one using a sodium LGS. With more photons available from the LGS, the sub-aperture size can be made smaller, increasing the degree of ground layer compensation and hence the degree of seeing improvement. The ultimate gain in resolution depends on the fraction of turbulence located in the lower layers and on the size of the desired FoV; for typical Pachón conditions it reaches a factor of 2.
In the IR, low order AO with NGSs provides roughly the same degree of compensation as on-axis tt, but in a much wider FoV. In the K band, a Strehl ratio of 0.2 can be achieved uniformly over a 4' field under median seeing, increasing to 0.3 for good seeing. If only on-axis compensation is needed, the Strehl ratios in the K band are 0.73 and 0.82 for median and good seeing respectively. In the intermediate case of a Rayleigh LGS at 10 km, a uniformly compensated FoV of 2' will have Strehl ratios in K band of 0.29 and 0.43 under median and good seeing.
Resolution gain in the visible remains modest, especially with NGSs. Low order compensation of the whole turbulent volume shrinks the PSF on-axis by the factor of 2, at most. Alternatively, high order compensation of ground layer provides a similar gain of 2. Trying to correct both low order and wide field gives smaller gains. At 0.5 micron, FWHM in a 4' FoV is improved by a factor of 1.22 with three 15-mag NGSs or 1.57 times with a Rayleigh LGS at 10km altitude. The gains in limiting magnitude (background-limited) are even lower, about 0.1mag and 0.15mag with NGS and LGS, respectively. The reason is that the shape of a partially compensated PSF is different from that of seeing limited PSF, in that the gain in central intensity and FWHM is larger than the gain in energy concentration.
The seeing improvement is far from being negligible, however. With Rayleigh LGS, FWHM of 0.35'' and 0.25'' can be achieved at 0.5 micron under median and good seeing, compared to 0.55'' and 0.40'' for uncompensated images. The information content of an astronomical image (number of independent samples) is tripled. In confusion-limited problems requiring moderately wide FoV, like stellar populations studies in nearby galaxies or globular clusters, SOAR with seeing improvement will beat any ground-based telescope and will approach HST performance.
The real gains will be larger in comparison with model calculations, as for tt compensation. AO will correct the telescope aberrations (including precise and automatic focusing) which were assumed to be absent in the simulations. The gains increase under good seeing, of course.
SOAR is already being provided with excellent optical and IR imagers with fine pixel scales and large FoVs. Images improved by AO should preferably be fed to these instruments rather than to additional specialized detectors forming part of an AO system. It implies that turbulence must be compensated with an adaptive secondary, because optical and IR imagers work at different focal stations (folded cassegrain and IR ISB). An alternative solution (a small DM with re-imaging optics) would mean that the imagers need to be re-located or duplicated; additional light losses in the re-imaging optics and increased emissivity in the IR will likely cancel the small gains in detectivity [2], leaving only the gain in resolution. Still, we propose a solution where an AO system with small DM feeds a dedicated optical imager with a FoV of 2'-3'.
 |
 |
| Figure 4: PSFs at 0.5 micron and median seeing for uncompensated imaging (left), on-axis, and at increasing distance from the guide star for a high-order AO with 66 Zernike modes compensated. | Figure 5: Strehl ratio on-axis and at 4'' offset (0.5 micron, median seeing) as a function of NGS R-magnitude for high-order AO. |
A classical AO system of moderately high order working in the visible range can achieve diffraction limited compensation on-axis, provided that a suitably bright NGS is available. By compensating 66 Zernike modes, a median Strehl ratio of 0.06 can be achieved at 0.5 micron, increasing to 0.16 under good seeing. At 0.8 micron, Strehl ratios are 0.35 and 0.52, respectively. Diffraction limited FWHM is 0.034" and 0.046" at 0.5 and 0.8 micron. Clearly, such a leap in resolution (3 times better than HST!) opens unique research opportunities, despite the very restrictive need of a bright NGS and a very narrow compensated FoV.
PSF variation over the FoV is illustrated in Fig. 4. In 10" FoV the changes are already dramatic. More uniformity can be achieved at the expense of reduced on-axis quality. It is also evident that, despite low Strehl ratios, the diffraction limited core is very prominent, favoring image restoration.
How faint a GS can be used? In Fig. 5 the Strehl ratio is plotted against GS magnitude. Note that with a bright GS, better Strehls will be achieved for compensation orders higher than 66. Alternatively, when the GS is too faint (e.g. 15mag), the compensation order will be reduced, the Strehl ratio will become very low, but some seeing improvement will still be obtained.
Rayleigh LGSs do not achieve diffraction limited compensation in the visible due to the cone effect. On the other hand, sodium LGSs are feasible and give almost the same performance as a NGS. We did not yet study the problem of tip-tilt compensation needed in conjunction with a sodium LGS, but initial estimates indicate that sky coverage will remain very low if diffraction limited performance is required. Alternatively, when the tt NGS is too faint or too distant from the target, the tt noise will enlarge the PSF core and degrade the resolution. This degraded resolution, however, may still be adequate for some programs. Given the high cost and complexity of sodium LGSs, the potential advantages of this upgrade must be carefully evaluated.
High order AO optimized for the visible and working at a 4m telescope will be without competition for a long time. Some programs, like astrometry of multiple stars (at present done with speckle techniques at the Blanco 4m) will benefit immediately. Studies of circumstellar matter using the IFU spectrometer will become possible at unprecedented resolution, with applications to stellar formation among others. Last but not least, measurements of relative positions of object features changing as a function of wavelength (chromatic position difference) will become possible with sub-milliarcsecond accuracy, giving access to the studies of inner accretion disks, envelopes of Be stars, jets, as will the measurement of dynamical parallaxes for distance determinations to expanding nebulae etc.
Diffraction-limited imaging in the visible would require a dedicated imager. A 1Kx1K CCD camera with 15 mas pixels covers a 15" FoV, well matched to the AO isoplanatic patch. Such an imager will be relatively straightforward to make as a part of the AO project. The IFU spectrometer can be easily adapted by providing fore-optics with suitable magnification. With 15 mas per fiber, the 26x50 entrance aperture of IFU projects onto 0.4x0.75" of sky.
If the high order AO is upgraded with a Rayleigh LGS, it can be adapted to the seeing improvement mode described above. In this case a dedicated imager with a pixel size of about 0.1" is needed. A single 2Kx2K CCD can cover a 3' FoV. This option seems to be a realistic alternative to the expensive AO with adaptive secondary.
 |
 |
 |
 |
The resolutions that can be achieved with the three options discussed above are compared here with uncompensated seeing limited image quality both with standard theory and with the finite outer scale taken into account. The parameters used in the simulations are listed in Table 2. The tt compensation with 3 NGSs is taken as representative (not on-axis tt with one NGS). A range of NGS magnitudes was studied. In the figures below, the performance for bright NGSs is given, to show atmospheric limitations, but for the faintest NGS magnitudes indicated in the Table reasonable performance can still be obtained. For the low order AO with seeing improvement we studied two options: NGSs in a 4' FoV and Rayleigh LGS optimized for a 2' FoV (and giving useful results in a wider FoV as well). In the figures below, "low order" refers to the LGS option of seeing improvement in a wide field, as opposed to high order on-axis compensation.
The FWHMs of uncompensated and compensated images as a function of wavelength are given in Fig. 6. The influence of a finite outer scale is very noticeable. The outer scale of 25m is typical for Cerro Pachón and other sites [8]. However, it varies considerably with time and may be either larger (e.g. 50-200m) or smaller (15m), so the actual atmospheric image size will be somewhere between the two curves. For a large outer scale, the amplitude of tip and tilt increases, hence the gain brought by tt and AO is larger.
High order AO gives diffraction limited FWHMs at all wavelengths (on-axis), although the Strehl ratio in the visible is very low. Naturally, the gain brought by low order AO is between those of tt and high order AO; the gain in FWHM at 0.5 micron is 35% for median seeing and 46% for good seeing.
Please, note that the FWHM of uncompensated images in the K band is already 0.25'' under good seeing and 0.33'' under median seeing; it improves at long wavelengths faster than wavelength to the -0.2 power due to outer scale effects. A telescope with perfect optics, without turbulence in the dome and without wind shake, should be able to achieve this performance.
The Strehl ratio is the most common metric of AO performance. In Fig. 7 (left) the equivalent characteristic, or Fried resolution, is plotted. It corresponds to the gain of the light intensity in the center of the PSF compared to the "standard" uncompensated atmospheric PSF (infinite telescope, infinite outer scale). The gain is expressed in magnitudes. In the right panel, the FWHE (Full Width at Half Energy) diameter is plotted.
Naturally, high order AO shows most spectacular gains, but these gains can only be achieved in the small field around bright stars.
As stated above, the need to serve several existing SOAR wide-field instruments and to maintain high throughput strongly favors an AO system with deformable secondary. In Fig. 8 a sketch of such AO system is given. Three NGSs are picked off at the periphery of the field and analyzed by three wave-front sensors (WFSs). When a Rayleigh LGS is added, projected from behind the secondary, these pick-off mirrors will still serve to measure tip-tilt on 3 NGSs, whereas returned laser UV radiation will be reflected by a dielectric coating (e.g. on one of the optical elements of SOAR imager) and directed to an LGS WFS. On the IR side, the optical and UV radiation will be reflected from the dichroic in the IR ISB and analyzed with 1-3 NGS tt sensors and LGS WFS, which will duplicate those of the optical imager.
The implementation of an adaptive secondary mirror presents a major technological challenge. An adaptive secondary was made for the 6.5 m MMT conversion and is planned for the LBT. The cost and complexity of these adaptive secondaries is high, and the delay for SOAR due to the commitment of this group to the LBT project would be at least 2 years. Starting an alternative project of adaptive secondary technology development will be costly and may be very risky.
If an AO system is built around a small commercially available DM, the project does not encounter technological problems and can be realized in a short time (1-2 years) and with a moderate budget. The optical concept of a high order AO system (Fig. 9) is based on the cheap electrostatic deformable mirror (DM) from OKO-Tech, Delft, Netherlands. Their 50mm model can have up to 119 actuators, the beam diameter is 35mm. The approximate cost of this DM including high voltage amplifiers and interface cards is $37K.
An f/16 beam going out from the optical ISB is focused in the plane (2). The quartz field lens (1) forms the pupil image on the DM (3) at a distance of 56 cm, thus the whole module is rather compact. Part of the slightly diverging beam corrected by the DM is reflected by a dichroic mirror (4) towards the WFS (9). The WFS will have a CCD as light detector (e.g. a CCD39 from EEV) and will be either a Shack-Hartmann or curvature type. In the latter case, both the intra- and extra-focal images will be projected onto the CCD simultaneously (no fast modulation) and read once per loop cycle, as in the S-H WFS. A provision will be made to work with only one extra-focal image, to further reduce the readout noise on faint objects.
Several interchangeable dichroics will be provided, to be selected according to the imaging wavelength of the scientific channel. The light path from the telescope focus to the WFS includes a minimal number of elements, assuring highly efficient use of stellar photons.
The turbulence simulator (10) with artificial star is permanently installed within the AO module and is placed into the beam for testing and adjustment. Turbulence is simulated by means of rotating phase screens, giving perfectly reproducible and well characterized distortions. The strength of simulated turbulence is selected by changing the distance of phase screen from the artificial light source.
If a LGS is used, one to three small pick-off mirrors or prisms are deployed close to the focal plane (2) to direct the light of 1-3 tt NGSs into the tt sensors (11). The tt compensation is achieved by the telescope tertiary mirror, hence there is no need to locate the tt sensors after DM, as in other AO systems.
After the dichroic, a steering mirror (5) directs the light beam to one of the scientific instruments mounted on the AO module. One of the instruments (shown in Fig. 9) is a high resolution CCD imager. It consists of a camera lens (6) (zoom lens), filter block (7) and 1Kx1K CCD detector (8) in a dewar. Small motions of the steering mirror enable some image scanning. This would be especially useful for the second instrument, the IFU spectrograph, because of its small sub-arcsecond field. The IFU fibers will be detached from their regular location and mounted on the AO module which will provide an image with a suitable scale.
The addition of more instruments (e.g. an IR imager) is possible, the AO module transmits IR radiation. However, should the IR AO be implemented, it would be wise to envisage a separate system optimized for the IR (low emissivity, special coatings) and serving existing IR instruments, instead of providing a special complement of costly IR instruments for the optical AO module.
It is possible to implement the seeing improvement option with this hardware. A 2Kx2K CCD with 0.1" pixels covers a 3' FoV (physical size 58 mm at the SOAR focal plane). A suitable pixel scale will be provided by replacing the camera lens. In this case a second plate scale for the IFU with larger pixels will be desirable too. The light of a Rayleigh LGS at 10km altitude, severely defocused, will be reflected again by the dichroic (4) and fed to the WFS (9), translated to accomodate for the defocusing.
This concept lends itself to progressive upgrades. As a first step, only the NGS option will be implemented, giving exclusive access to problems related to diffraction limited studies of bright stars and their immediate surroundings. At the second stage, a Rayleigh LGS and tt sensors will be added, accompanied by the wide field re-imaging lens to change the CCD plate scale. In this configuration, AO will give access to a wide class of stellar population and extragalactic studies, originally planned for the SOAR optical imager. Further along, the addition of a sodium LGS may be envisaged (if by then the technology of sodium lasers has matured enough) and new instruments can be added. The AO module can be quickly upgraded following any new revolutionary ideas likely to appear in adaptive optics (e.g. new principles of wave front sensing).
 |
| Figure 11: The options for SOAR AO are drawn schematically as rectangles with lower limit corresponding to resolution and right limit to FoV size. The length of the rectangles is related to the total number of pixels in the delivered images. |  |
In Fig. 10 a rough summary of AO options and their performance is presented. Links to scientific programs are very preliminary and certainly not exhaustive.
The low order (seeing improvement) AO serving the existing optical and IR instruments is scientifically compelling. However, this option would be efficient only with an adaptive secondary and a Rayleigh LGS. Very rough estimates of cost and implementation time are $2M and 3-4 years.
High order AO with natural guide stars is relatively straightforward and cheap. It can be implemented in 1-2 years at a cost of $1M or less. The major cost component is likely to be the software. The two instruments served by high order AO will be a high resolution imager and the IFU spectrometer. A third instrument port may be provided for future upgrades (e.g. a low resolution spectrometer).
Within the same concept of high order AO, it is possible to implement a seeing improvement capability with a moderately large 3' FoV. Such a system must be equipped with a dedicated 2Kx2K CCD imager and a Rayleigh LGS. By changing the LGS altitude, some tradeoff between compensation quality and PSF uniformity is possible, adding more flexibility.
Send comments to: Andrei Tokovinin
