LBT FACILITY INSTRUMENTATION PROPOSAL

Authors: A. Richichi*, C. Baffa*, G. Brusa*, A. Cimatti*, C. Del Vecchio*, S. Esposito*, L. Fini*, S. Gennari*, F. Lisi, R. Maiolino*, A. Riccardi*, P. Salinari*

R. Ragazzoni**,..,..,

M. Busso***,..,...,

*) Osservatorio Astrofisico di Arcetri

**) Osservatorio Astronomico di Padova

***) Osservatorio Astronomico di Torino

More than a single and specific instrument we are here proposing to develop a family of instruments devoted to exploit the full potential of LBT for high angular resolution work at high sensitivity. In order to obtain the ultimate performances a considerable development of the adaptive optics system is necessary.

We will discuss here only the conceptual grounds on which we base our proposal. The work to evaluate in greater detail the instrumental options and the potential cost of the proposed interferometer and of the associated advanced Adaptive Optics system is still to be done, and will certainly require a number of years and the involvement of many collaborating Institutes.

In the following we will try to discuss various options for this ambitious program, balancing scientific wishes with our present understanding of technical possibilities, but we are fully aware of the tentative nature of our present preliminary discussion.

The reasons of presenting this proposal at the present time are the following:

• We believe that in the first decade of next century instruments similar to the ones we describe here will be able to provide a unique and essential complement, both in imaging and in spectroscopy, to space born astronomy, including not only HST but also NGST.
• Consequently we think that a considerable part of the LBT instrumentation budget has to be reserved to interferometry at short wavelength, and the present Call for Proposal is therefore the appropriate occasion for discussing a preliminary budget allocation.
• A large fraction of the investment, both in terms of money and of work, required by the high angular resolution instruments consists in developing an adaptive optics system considerably more advanced than the one already included in the telescope budget. Various other instruments can take advantage of this investment on AO, while some of the modes described here as options that can be implemented at the combined focus could also be part of instrument at different foci. This is therefore an appropriate time for optimizing the instrument complement of LBT.

This is not a "private" proposal. We look for the involvement of many of the groups participating in LBT in a coordinated program aiming to make of LBT the powerful instrument it can be. The work we propose is challenging and vast, only a large participation of LBT partners can accomplish it in a reasonable time. Of course we desire to contribute work in this program in areas that will have to be agreed.

As mentioned above, the key point of this proposal is that we believe that a further dramatic step can be done in adaptive optics, pushing this technique to very high quality correction over a large fraction of the sky even at optical wavelength. In order to achieve this ambitious objective a number of techniques have to work together. The first section of this proposal will therefore deal with the conceptual approach to the advanced adaptive optics we need. The second section will discuss instrumental options and priorities, the third will attempt to indicate a possible configuration for the combined focus and for the instruments. Section 4 reports preliminary ideas on an implementation sequence, section 5 attempts a census of required expertise areas, and section 6 reports very rough cost estimates for at least a part of the necessary developments. Section briefly 7 reports scientific priorities and possible areas of involvement of the proponents. 

The LBT baseline AO system is already fairly advanced and includes technology that is still under development, such as Sodium lasers and adaptive secondary mirrors. D. Sandler and R. Angel [1] have shown that an adaptive system such as the one envisaged for LBT, with a single artificial star and a single corrector, can achieve excellent correction (Strehl ratio of about 0.5 in the K band) over a large fraction of the sky.

Both, quality of correction and sky coverage drop unfortunately going to shorter wavelength if only a single artificial star is used for correction, due to focus anisoplanatism. Although the use of bright natural stars allows excellent correction even at optical wavelength, the resulting sky coverage is extremely low. Removing focus anisoplanatism is therefore one of the key factors for extending AO to shorter wavelength with a significant sky coverage.

In a paper of several years ago Tallon and Foy [2, TF] have shown that by using a modest number of artificial stars the focus anisoplanatism effect can be totally removed. Moreover the Tallon and Foy multi-laser technique makes also possible to obtain turbulence tomography, separating the contributions to the instantaneous wave front phase error that are due to different layers of the atmosphere. Although not yet tested in practice, for the obvious reason that even single laser AO is still in its infancy, the TF method is essentially based on geometry and is therefore fairly solid if laser stars work. Extensive simulations of this method are presently in preparation in the frame of a European TMR program.

With the TF technique it becomes therefore possible not only to correct high orders of the wave front error, but also to apply separate corrections for different atmospheric layers. The consequence is that one can use "multi-conjugate AO", a technique described by J. Beckers [3], that extends considerably the isoplanatic angle. The combination of the two techniques therefore provides the possibility of correcting the wave front not only without the limit posed by focus anisoplanatism, but also over a much larger field of view.

Table 1 shows the Strehl ratio of the dominant residual error, the "fitting error" in the above assumptions as calculated from a model of the adaptive secondary mirror. Fig. 1 shows a section of the corresponding PSF. Although a significant fraction of the energy is missing from the central peak, the residual wave front error is on small spatial scales, therefore diffracts energy on large angular scales so that the contrast between the peak and the scattered light is extremely high, >10³. This is an important feature for interferometry, where the reduction of the fringe contrast caused by scattered light is negligible.

Table 1: Strehl ratios corresponding to the dominant error term in a multi-laser double corrector system with the parameters reported in the text.

We will now go through the various components of the multi-laser, multi-conjugate AO system just

outlined to understand qualitatively what we need. Concerning the laser system we refer to [2] for a description of the geometry, of the algorithms and of the capabilities of the multi-laser technique, and we only answer a few simple practical questions:

Based on the analysis of Tallon and Foy the minimum number of lasers that accommodates our preliminary field requirements is four per pupil. The maximum field angle over which a full correction is possible in this case is about 50 arcsec, while it would be too small (~7 arcsec) with 3 lasers. The four lasers would have to be projected at an angle of 83 arcsec from the telescope axis.

The requirements on power are somewhat less than for a single laser systems. Lasers currently under development for single laser systems are therefore more than adequate.

The laser beams can be projected exactly like in the single laser system we have foreseen, from the back of the secondary. Each artificial star will then be analyzed by a "normal" wave front sensor, for instance a Shack Hartman sensor with about the same spatial resolution envisaged for single-laser systems. The inclined laser beams will produce a much stronger background of scattered sodium light close to the artificial star images, and an efficient baffling system has to be provided.

None, if we assume that the laser technology will be available in the next few years. Certainly the reliability and cost requirements are different from those of a single laser. The current development at University of Arizona seems to fit all requirements including moderate unitary cost (~250 k$ per laser).

The proposed scheme of double-conjugated adaptive correction requires two correctors for each beam, conjugated to suitably different heights in the atmosphere.

The first corrector is obviously the Gregorian adaptive secondary, whose conjugated plain lays about 100 m above the primary. The secondary is undersized to be the pupil, determining an effective telescope aperture of 8.25 m. The actuator spacing will be between 25 and 30 mm, corresponding to 25 to 30 cm on the wave front, as required for use as single corrector during initial phases and for other focal stations where double-conjugate AO will not be implemented.

The basic technology of the adaptive secondary has been already developed and is currently adopted to construct the first telescope unit, the MMT F/15 adaptive secondary. An important feature of this device is that an internal position loop controls the position of each actuator. It is therefore possible to operate the corrector by updating "absolute" actuator positions rather than zeroing the residual error on a star. This is essential when using multiple laser stars and multi-conjugated correction, because in this case the control loop cannot be closed directly on a star in the usual way.

In addition to correct the high orders for low atmospheric layers, the adaptive secondary will correct the wave front inclination produced by the entire atmosphere and in the optical path inside the telescope and the instrument. Using a single corrector for tip-tilt causes negligible de-correlation of the higher orders. In our configuration, where the inclination is corrected at the low conjugate, the de-correlation on the highest layers (H~10 km) is <1 cm, while the value of r_o for a high layer in reasonably good seeing conditions is likely to be more than 30 cm in V.

We have preliminarily explored a few optical layouts for the second stage of correction:

1. a spherical corrector illuminated by the telescope F/15 beam (two or three added reflections)
2. a double pass off-axis parabolic collimator illuminating a flat corrector (four added reflections).
3. a corrector coincident with the beam combiner flat mirror (no extra optics).

Although configuration a) and b) give more freedom in choosing the conjugate position, the optical performances where unacceptable at a first attempt, due to the large off-axis angles required in order to avoid vignetting of the beams. The optical efficiency of the first two configurations is also not ideal at visible wavelength due to the extra reflections. Further work could identify better solutions, what counts for the moment is that at least configuration c) seems to be optically viable, and, in fact, very interesting. We will therefore continue the present qualitative discussion of the corrector requirement with reference to the "correcting combiner" (CC).

If the beam combiner coincides with the corrector there is no additional optics to be added, and the fast and small amplitude optical path difference corrections can be done in an optimum way by the correctors. Slower and larger amplitude compensation of optical path difference, due for instance to thermal and gravity deformations of the telescope structure, can be done by displacing the entire CC unit, like with a passive combiner.

Of course placing the second corrector at the beam combining flats introduces some constraints, but fortunately these don't seem to limit seriously the performances. If we don't want to introduce extra reflections, the position, and therefore the size, of the second corrector are determined by the available space (less than 2 m at the central combined focus). Preliminary optical design shows that in this case a simple doublet of ~18 cm diameter can form on the beam combiner flat an image of a layer located about 6 km above the primary (~9 km above sea level). This is an excellent conjugate position for observations within about 40 degrees from zenith, where the height of the conjugate would become ~4.6 Km (~7.6 Km above sea level).

The on-axis image of the layer formed on the flat combiner mirror has a diameter of about 80 mm. The doublet has excellent Strehl ratio over about 2x2 arcmin. Of course the beam combiner must be at 45 degrees inclination, therefore the conjugates of opposite edges of the corrector are at a different height. The conjugate plane is inclined, going from about 5.5 to about 6.5 km at the opposite edges of the beam. This means that even if we had only a single turbulence layer exactly at 6 km, there would be a finite isoplanatic angle due to the change in conjugation height across the beam. In most practical circumstances this effect is likely to be negligible.

Although the basic technology for building the CC is (nearly) available, there is no doubt that developing this device is one of the challenging aspects of our proposal. The system has about the same complexity of the two adaptive secondary mirrors and is likely to be about as expensive. On the other hand it can provide the best possible optical efficiency (no extra optics for the second stage of correction, exactly like the adaptive secondary for the first stage!) and can make the differential piston correction extremely fast and accurate.

There is a serious drawback in adopting double-conjugated AO: it is really difficult to cool down the entire instrument to cryogenical temperature, in particular because of the second corrector. This is not a real problem for R, I, J, and H, but limits the achievable sensitivity in K to about 1/2 magnitude less than the theoretical limit for a cold system. The possibility of cooling the CC to approximately -60 degrees C, sufficient for recovering the maximum sensitivity in K, has to be examined carefully in deciding on CC technology.

The CC is not only the second stage of adaptive correction, but can also compensate optical path differences and pupil geometry variations of the two interfering telescopes. The first compensation can be performed by translating rigidly the entire unit, the second by small changes of the mirror spacing.

Pupil geometry must be preserved with a relative accuracy of the order of a few PpM if we want to preserve astrometric accuracy over the interferometric field. The primary mirror spacing is subject to changes of about 10 PpM per ⁰C, due to thermal expansion of the steel structure and to changes due to gravity deformation of the telescope which are of a fraction of a mm between zenith and horizon pointing. Both can be corrected open loop.

Although in principle no special developments are necessary for sensing high and low orders when using a Tallon and Foy multiple star system, we will briefly discuss here some aspects connected with our proposed scheme.

The separation of the laser beams from the object beam is conveniently feasible in proximity of the bent Gregorian focus. The laser star spots could be easily reflected off the main beam by small reflectors, as they are more than 80 arcsec off axis. Depending on optical efficiency in the science bands of Sodium doublet reflectors it might be convenient to reflect off the main beam all the Sodium light coming from the artificial stars and from lower atmospheric back scatter.

The unshielded out of focus Sodium light is in fact several orders of magnitude more intense than the broad band natural sky background in the V band, and has to be suppressed very efficiently when working with detectors sensitive at 589 nm and on extremely faint sources. Sodium stray light can even reduce the sensitivity of the high order wave front sensors, that also must be shielded, although they are well decoupled from most of the Raileigh light and can tolerate a much higher stray light flux.

A convenient option for baffling most of the stray sodium light is offered by the Gregorian configuration of LBT. A suitable field stop, with apertures for the science beams and for the laser stars at the primary focus can remove most of the out of focus Sodium light produced in the low atmosphere.

The consequence of separating the laser beams in front of the instrument is of course that we cannot measure high order wave front error produced within the instrument with the lasers. We therefore have to take care of reducing as much as possible "internal seeing" in the instrument, avoiding the formation of turbulence on scales comparable or smaller than the beam cross-section (~100 mm). An ideal system would be that of evacuating the entire instrument. This would certainly be the preferred option if even a moderate cooling of the CC is possible. A simpler alternative is to enclose the instrument in a sealed thermal protection and fill it with Helium at atmospheric pressure. The ~ 12 cm (minor axis) laser beam splitter would act in both cases as instrument windows, but without differential pressure if the system is filled with Helium. Within the Helium filled instrument a temperature uniformity of a fraction of degree C has to be achieved to effectively prevent "instrument seeing".

The low order sensing must be done after the second stage of correction to profit of the wider corrected field. It is difficult to imagine a metrology system with the same level of completeness and reliability as the beam from a natural star that goes through most of the optical system with a full beam. Sensing the low orders as close as possible to the final images is therefore a good way of correcting errors due to mechanical flexures, thermal expansion and refraction index variations during observations. Of course a "system test mode" with higher order sensing on brighter stars can be used several times, if needed between observations.

We need to sense at least five quantities, tip and tilt for the individual beams and differential piston. Whether we need to sense other low order errors is a crucial question that will have a clear answer only after a sufficient experience with the use of single laser AO systems. The main question concerns the focus term. The artificial stars are produced by resonant scattering in a layer that is > 10 km thick, therefore any change in the Sodium height distribution in the layer affects the focus of the laser stars. In a binocular telescope like LBT the Sodium vertical distribution can in principle be monitored continuously without the need of extra stars or extra telescopes, because the laser stars projected from one of the telescopes are significantly elongated (several arcsec) when seen from the other. As the laser spot are separated by typically 40 m on the Sodium layer, even small variations of Sodium equivalent height on smaller spatial scales might force to use a natural star for focus correction. This of course has a cost in photons and therefore in sky coverage.

The four degrees of freedom associated with tip-tilt are more difficult to sense than differential piston because they change at higher frequency (roughly by a factor of two) and because they only use the photons from one of the beams. The focus term, if not provided by the laser stars for the reasons discussed above, is somewhat less demanding than tip-tilt because it is only one degree of freedom and it has less power, even if its rate of change could be somewhat higher than that of tip tilt. The previous considerations on sky coverage wouldn't therefore be severely affected by the need of measuring focus.

The type of interferometry that can be done with LBT is unique both at short and at long wavelengths, and the first question to answer is therefore about the spectral coverage. We are convinced that both spectral regions (NIR, with extension to I and R, and, if possible V, and the Middle IR) must have high priority, but it is extremely difficult to combine the entire range of wavelengths in a single instrument. The question is therefore how to divide the spectral range in an optimum way, taking into account efficiency, cost and risk.

The two instruments are highly complementary for many fields of astronomical research and should both be available as common user instruments in parallel, therefore will have to be designed for two different focal stations. The natural division, dictated by the very different sensitivity to differential retardation induced by the tertiary and beam combiner flats, is to have the LWS at the rear beam combined focus and the SWI at the central one. Both instruments need adaptive optics, although at a different level of complexity, and both must have operational modes that could produce significant science even with reduced or missing adaptive correction. LWS could be forced in this mode by a delay of the baseline adaptive optics system, while SWI should work, at the beginning, without the advanced adaptive correction that will be discussed in the following.

Assuming the above division of wavelength ranges between the two Interferometers we now go through a list of options that could be considered for the SWI

Interferometric imaging at the highest angular resolution and sensitivity in each band is clearly the prime scope of the LBT Interferometers. This is not a simple task, especially at short wavelength, but is one that could make LBT unique not only in comparison with other ground based telescopes of similar size but also in comparison with HST and NGST. In particular at optical wavelengths the atmospheric background is not dramatically higher than in space, and the possibility of competing with space borne instrument is based essentially on the quality of the adaptive correction of the effects of the atmospheric turbulence. This is the main reason why we believe that achieving excellent correction over at least part of the optical domain is crucial for the scientific excellence of this instrument. In the previous section we reported our considerations on how to obtain the desired adaptive correction. Here we will describe what we wish to obtain in imaging.

Table 2 reports single pupil and interferometric angular resolution, angular pixel sizes and F/n of the individual beams, based on assumed pixel sizes. We expect we could correct a field of approximately 30 arcsec diameter with high Strehl ratio in the R band. A look at Table 2 shows that we need about 20,000 by 20,000 pixels to cover that field in R in interferometric mode, if we sample at the minimum possible sampling to avoid loss of fringe contrast, ~ 4 pixels per fringe.

Table 2: Angular resolution, angular pixel size, field and beam numerical aperture for single pupil and for interferometric imaging. The telescope effective diameter D is 8.25 m and the max baseline b is 22.65 m. All F/n refer to single pupil beams.

The number of pixels necessary to cover the well corrected interferometric field in the other bands is of the same order, due to the increase of the corrected field at longer wavelengths. The detectors are therefore likely to be a significant part of the instrument cost, although it is certainly possible to start with a single chip and a reduced field, as reported in Table 2. A single InSb array could in principle cover the entire range R-K, but this solution is not advisable for various reasons (first of all sensitivity to thermal background). We consider as baseline choice that of (at least) one 4000x4000 CCD for R and I and of a 2000x2000 HgCdTe detector for J, H and K.

Excellent, but not impossible, performances in readout noise and dark current are necessary in R and I, where, as shown in Table 3, interferometric imaging could be easily limited by detector performances. In J, H and K the background is higher and the detectors, using multiple readout during integration, can more easily reach the required performances. In the K band we have assumed a background ~0.8 mag brighter than the sky background to account for the contribution of ambient temperature optics.

Table 3: Noise equivalent magnitude per pixel for the case of interferometric imaging. Assumptions are: 1000 s of integration time, 30% transmission for both, background and signal photons.

The sensitivity achievable on unresolved or barely resolved sources is affected by many parameters, first of all the quality of the adaptive correction and of the co-phasing.

In case of perfect optics, correction and co-phasing, ~80% of the photons from an unresolved source would fall within the first zero of the Airy function of a single pupil, corresponding to about 550 pixels at the sampling adopted in Table 3. About one half of this light is concentrated on about 100 pixels. An average signal to noise ratio of 10 in the fringes could therefore be obtained in only 1000 seconds on point sources about 6.5 mag brighter than the noise equivalent magnitudes (per pixel) reported in Table 3. At this level of S/N the interference pattern would be clear enough to allow PSF de-convolution algorithms to derive full two-dimensional resolution and better S/N from multiple exposures.

It is clear that the about six magnitudes of difference in sensitivity and the factor of about 2 in angular resolution between R and H justify the effort of pushing the interferometer to the shortest possible wavelength. Red-shifted ultraviolet and optical features of high z galaxies are certainly a prime candidate for interferometric observations in R and I.

This is the basic mode with an adaptively corrected single pupil, and can be obtained in the combined focal plane simply not applying the phase correction, or physically displacing apart the two images on the same detector or on different detectors. In all cases the sensitivity gain is about the same that can be obtained by degrading of the same factor the angular resolution in interferometric imaging by binning the pixels. The difference is of course that one can cover a much larger field in this mode if the detector is matched to a lower angular resolution.

This is only another name for Speckle Interferometry. A variety of techniques allow to recover high angular resolution information and some of them can reconstruct complex images. The common requirement is that of freezing the effect of the atmospheric turbulence by using very short integration times. It is clear from table 2 and 3 that in very short integrations (<0.1 s) the detector noise determines the sensitivity, that is therefore greatly reduced. On the other end these techniques can be used in a total absence of adaptive correction or with partial adaptive correction and no co-phasing. This makes of Speckle interferometry a very valuable tool for early phases of the telescope life, when only moderate adaptive correction will be available, and for later phases in areas of the sky where natural guide stars are not available. In this case one can take full advantage of the high order correction and this "single speckle" case becomes another way of calling the "shift and add" method.

From a technical point of view the only specific requirement different from those of the other imaging modes is that of a fast readout of the detectors. This is easily obtainable with existing IR detectors, where only a subset of the detector can be read at correspondingly higher rate, while is not yet common on optical CCD detectors, but optical detectors with addressable readout, similar to those uses in the NIR, are currently in development.

Although images are of great importance in understanding astrophysical phenomena, a quantitative study normally requires spectral information. This is in principle obtainable at the same angular resolution as interferometric images, but of course at a significant cost in sensitivity. Nevertheless for sufficiently bright sources this provides a very rich and detailed information. For faint resolved sources it is convenient to increase the slit width to collect more photons, because it is clear from Table 3 and 4 that even at a modest spectral resolution the sky background is not going to be the dominant source of noise except in the K band. In fact Tables 3 and 4 cannot be scaled easily with respect to spectral resolution, because the background is largely concentrated in strong sky emission lines, and it is therefore much lower over most of the spectrum than average values scaled from the tables. Detector readout noise, dark current and cosmic ray events will determine the optimum integration time and sensitivity in most cases. We will not go through all possible combinations of angular and spectral resolution at the various wavelengths, but only discuss a few basic facts that are common to the various combinations.

The entire spectroscopic end of the instrument, from the slit to the detector, has to be de-rotated, and again the maximum duration of each integration is determined by PSF rotation in a similar way as in interferometric imaging.

Over most of the spectral range considered here detector noise limited sensitivities should be obtained for slit widths of the order of 100 mas. An example of this mode is reported in Table 5. Here it must be noted that the effective resolving power is ~500, one half of the value reported in the third column that refers to the spectral coverage of a single pixel. The values of the background suppression factors, in the fifth column, are only very crude estimates. The last column gives the magnitude of a source filling a pixel whose continuum could be detected with S/N =1 in the usual assumptions for integration time (1000 s) efficiency etc.

With longer integration times (several hours) one could therefore detect the continuum of compact sources (~ 0.1 arcsec) at about mag 30 in R and I. Even more spectacular are the performances in the NIR, due to the OH suppression. This spectroscopic mode would then give access to spectroscopy of essentially all the sources in the Hubble Deep Field, where, by the way, there seems to be a cutoff right at angular sizes of about 0.1 arcsec.

Table 5: Example of a "wide slit" combined, but not co-phased, spectroscopic configuration. The last column reports the magnitude, integrated over the selected pixel, detected at S/N =1 in the continuum. Both, background and signal are calculated with the full telescope aperture (two mirrors). Other parameters are as in Table 3.

A very large telescope with very good adaptive correction in principle is an ideal instrument for extremely high resolution work. The main advantage of the adaptive correction is that a fairly compact spectrograph can provide an extremely high resolving power on a nearly diffraction limited slit. Moreover in most cases the source itself can be used for low order correction, so that the sky coverage is essentially complete. We didn't even attempt a conceptual design of such an instrument, which differs significantly from the other low dispersion spectrographs discussed. On the other hand an optical layout compatible with the proposed configuration of the corrected-combined focus doesn't seem to be impossible even if cross dispersion is needed.

We spend here only a few words on more specialized instrumental modes that could greatly profit of the advanced correction and/or of the binocular nature of the telescope and that seem to be compatible with the general layout we assumed for the imaging and spectroscopic modes.

This mode is likely to be implemented in the LWI specifically to study the zodiacal light emission in the thermal IR of nearby stars, an important parameter for planning future space missions devoted to the search of earth-like planets. In principle there is no problem in implementing nulling interferometry at shorter wavelength in the SWI. The nulled field would be reduced in proportion, while the light "leaks" would be larger for various reasons, but still this way of implementing a coronographic mode is likely to be much more effective than conventional coronography. The implementation of this mode is conceptually simple, because in principle it is sufficient to pick up the two beams after the combiner/corrector or even after the field re-imaging optics and make them cross at the position of a beam-splitter.

Although a non negligible amount of linear polarization is certainly introduced by the two reflections at 45 degrees on the tertiary and on the combiner, there are various ways of obtaining accurate linear polarization measurements at high angular resolution in all imaging modes. In non co-phased imaging modes a fixed bi-refringent prism can be used, for instance, in the converging beam and the telescope rotation could be used to explore 90 degrees. The signal of a field star (we always have in the isoplanatic field at least a "bright" one for adaptive low order correction) could be used to remove transmission variations.

In the interferometric (co-phased) imaging mode the double modulation (fringe rotation and polarization angle) might be difficult to disentangle and it could be better to rotate the polarizer in a number of positions for each interferometric sub-integration. Several images at different polarizer angles would then be reconstructed and compared.

Similar consideration can also be extended to spectro-polarimetry. Although these considerations are extremely rough, we want to stress the importance of considering a high angular resolution polarimeter option

We have considered a variety of observing modes at the combined LBT focus and an advanced adaptive optics system. The problem is now to find a way of satisfying the contrasting requirements of the various components for all, or at least for most, of the observing modes we described. The basic requirements emerging from the above discussion of observing modes and of adaptive requirements are the following:

1. We need to form an image of a high atmospheric layer on a deformable mirror
2. We need to image the field on the detectors at very different scales (from about F/6 to about F/200)
3. We need to have an intermediate image of the field for spectroscopy (diffraction limited slits diffract!)
4. We need to find appropriate positions for two beam splitters (for laser and natural star wave front sensing)
5. We need to de-rotate detectors and spectrographs
6. We have to build up the instrumentation we discussed gradually, starting from the simplest configurations.

Work done in the past on a similar configuration by P. Bayard and D. Bonaccini [4] has shown that relatively simple transmission optics (spherical doublets) can provide a very high Strehl ratio over a wide field and a very extended wavelength range, provided re-focussing is foreseen for the different bands. Although very little optical design has been done yet to verify the actual feasibility of this instrument, we have adopted a layout entirely based on transmission optics, sketched in Fig.3, because we expect it to represent a viable solution. Of course different alternatives will have to be explored, especially for the field re-imaging optics

In the scheme represented in Fig. 3 the re-imaged focal plane acts as a new focal station of the telescope, where the beams are already adaptively corrected and combined, in a similar way to the Gregorian foci for single beam, single adaptive correction. The list of modes of operation we have presented in section 2 could be considered as a list of possible instruments for this new focal station.

Some of the instruments do not need an intermediate re-imaging of the field, so that an obvious alternative scheme is to consider "instrument" whatever is after the CC. In fact it is likely that even the collimator could be changed to optimize optical quality and transmission, depending on wavelength and application. The complex and expensive CC, the laser star sensors, the optical support structure and the "instrument de-rotator" are the components that need to be designed for common use of all the instruments. In Fig. 2. the de-rotator is placed not far from the pupil image to provide the option of intercepting the parallel beams in the "instrument" at the most convenient position.

In most cases the low order sensing is done at the same plate-scale as the science object, and therefore no much other optics is needed if the instrument optics has a sufficiently wide field. In some cases, a typical example is that of Phased Spectroscopy with OH suppression, the image scale is too fast for low order sensing and a separate optics is needed for this purpose. This can either be arranged in the Instrument or after the beam combiners in the parallel beam by inserting there appropriate beam-splitters (no co-phasing is required in this case).

Table 7: Results of preliminary optical analysis of a doublet collimator

The image quality at the CC is adequate if we consider that the corrector has elements of size ~ 3x5 mm. The collimator described in Table 7 is the one reported in Fig. 2.

The two pupil images, 80 mm in diameter, are separated by about 60 mm to preserve pupil geometry. If we allow for about 1 arcmin field, the field re-imaging optics (presumably another doublet, or a triplet, that we will call "camera" in the following) would be ~ 280 mm X 120 mm if placed in the pupil image plane. The camera doublet represented in Fig. 2 is only for illustration of the concept and has not been designed, as a number of different cameras are likely to be needed for different instruments.

The implementation of the above capabilities must take into account a number of constraints including the fact that in an initial period of about two years only one of the primary mirrors will be available and the various stages of the evolution of the adaptive system. Table 7 summarizes the instrumental options available at different stages of adaptive optics. Each cell reports the bands in which useful work can be done and the telescope configuration that can be used.

Table 7: Bands in which the main observing modes can be used at different stages of the development of the adaptive system. Of course there are significant differences in terms of accuracy, sensitivity and sky coverage for the same band and mode depending on adaptive configuration. "Ngs" and "lgs" stand, respectively, for laser and natural guide star, "corr" stands for corrector. The numbers in brackets refer to the telescope configuration: 1 stands for single primary, 2 for the binocular configuration. See text for further comments.

Apart from the obvious need of a binocular configuration for all interferometric work, all the other instruments can work in the NIR (and in some cases also in the "optical" range), although at reduced performances and sky coverage, with a single mirror and in essentially all the adaptive configurations. This means that we can adopt a step by step evolutionary scheme that reduces both risk and cost. The possible steps are the following:

In each of the above steps one has the possibility of producing valuable science and of testing subsystems that are needed for the further step. In step 1 the laser system is tested before multiplying the number of lasers, in step 2 the multiple laser system can be used for eliminating focus anisoplanatism problems, but will provide real data to be used for optimization of the double conjugate scheme. In a similar way, as soon as the second telescope becomes operative one can start testing and debugging the interferometric mode.

Clearly the above program overlaps in part with the program for other instruments, in particular with the diffraction limited modes of operation of the Near IR Spectrograph/camera. On the other end this can be used to divide functions among instruments, making all of them simpler.

Although the precise time sequence is difficult to define at the present pre-conceptual stage, the goal of having the entire multi-laser/double-conjugate/multi-instrument system available and working soon after the installation of the second mirror is not in principle impossible. This will depend from a good coordination of the activities more than from technical or financial problems.

Even a preliminary feasibility study of the many aspects of this proposal represents a massive work that requires activity of different groups with different expertise. We simply list here in random order a number of activities that, at a preliminary study level, are sufficiently independent from each other to make it possible to divide them among different groups.

1. Characterization of the turbulence parameters at Mt. Graham. The main goal is to obtain some statistics around the year on the vertical distribution of the turbulence.
2. Exploratory optical design of the entire system including a number of instruments. The scope should be, in this phase, to check the compatibility of the parameters of the combined focus (or of the a-focal combination) with those of the various instruments.
3. Comparison of various techniques for wave front sensing for low (includes piston) and for high orders
4. Simulations of multi-laser, double-conjugate systems.
5. Optical model of the entire interferometer and interferometric tolerance analysis.
6. Exploratory mechanical (and thermal) design of various options (vacuum at ambient or low temperature, Helium at controlled temperature) for the combined focal station.
7. Preliminary study of the CC.
8. Preliminary study of algorithms for best extraction of information from the various modes of operation.
9. Study of Sodium light notch filters and reflectors and, in general, of stray light suppression.

The above list is certainly not complete and doesn't include activities that are already going on such as the development of the lasers and that of the adaptive secondary mirrors.

It is only after the completion of this study phase, possibly in about one year, that a well defined plan can be formulated and precise responsibilities assigned for the next phase of design and construction of the various sub-systems and instruments.

There is certainly no precise answer to this question at this time. Still it is useful to attempt to define reasonable goals.

1. The Lasers: the current development at UoA is aimed to produce Sodium lasers at 250 K$ each. Taking into account the large number we need (8-2(baseline)+1(spare)=7), that the cost is likely to go down with time and that the projection optics is already in the baseline AO system, we can assume that the cost of the upgrade from 2 to 8 lasers will be < 2 M$.
2. The Correcting Combiner: It cannot be more expensive than the two adaptive secondary mirrors, because the optics is much easier to produce and the mechanics is much simpler. The electronics could be essentially the same, while the actuators (most likely piezos) could be somewhat more expensive. A cost of about 3/4 of that of the two secondary mirrors, therefore ~1.5 M$, seems to be a reasonable estimate.
3. Optics and Mechanics of the Combined Focus: the rest of the hardware necessary for beam combination, including the 8 laser wave front sensors and their detectors, the collimators, one of the cameras and the instrument de-rotator is likely to cost about 1 M$.
4. Instruments: these can become very expensive if one wants to exploit the enormous field we can provide in principle. For (relatively) small field instruments based on a single large chip the average cost of the instrument can be low, considering the relatively simple optics and mechanics. Including the low order wave front sensors an average cost of about 1/2 M$ per instrument (here this means per function, or, per mode, as more than one of the modes of operation could be implemented in a single instrument) doesn't seem to be out of reality.

We first of all reaffirm our desire to collaborate with our LBT partners in all phases of the development of the high angular resolution potential of LBT, starting from conceptual definition and division of work in the preliminary phase. The Italian community has obviously interest in essentially all the instruments that we have mentioned, as it is probably the case for our non-Italian LBT partners. The following considerations are therefore only a preliminary attempt to identify priorities in scientific interest and areas of technical experience among the group of Institutes presenting this proposal.

7.1 Science priorities (To be completed after preliminary distribution)

7.2 Technical expertise (To be completed after preliminary distribution)

References

[1] D. Sandler and R. Angel "Citare"

[2] M. Tallon and R. Foy "Adaptive telescope with laser probe: isoplanatism and cone effect"

Astron. Astrophys. 235, 549-557 (1990)

[3] J. M. Beckers "Increasing the Size of the Isoplanatic Patch with Multiconjugate Adaptive Optics"

in eso Conf. on "Very Large Telescopes and their Instrumentation", M. H. Ulrich ed., p 693 (1988)

[4] P. Bayard and D. Bonaccini "Citare"