(1) Smart Structures and the management of deformations
The vast and partially-exposed structure of an ELT will need to be able to correct, at high bandwidth, for both solid-body and internal perturbations, both of which will be inevitable consequences of exposure to wind. The OWL and MAXAT teams have naturally taken these issues particularly seriously, as neither can realistically expect to house their vast brainchildren in permanently protective structures (which have been assumed in their baseline designs by proponents of the smaller CELT and super-HET concepts). Considerable discussion is available in the literature and in circulation: Ardeberg et al.1996 discuss in detail the deformations of a ``point design" for a 25-m facility; Brunetto, Koch & Quattri (1999) examine, inter alia, the eigenmodes of a point design for OWL. The predicted overall behaviour of the current design for OWL is remarkably encouraging: mechanical modelling predicts locked-rotor eigenfrequencies close to those of many of today's 4-m telescopes.
This issue may overlap into the management of the optical system for AO purposes, as the frequency and actuation demands are not too dissimilar: the same effective precison is required of the telescope structure and of its optical components, which will be subjected to the same disturbances (variable wind loads). It seems certain that sophisticated laser metrology systems will be needed to determine and monitor the corrections required, while high-bandwith actuators of a variety of dynamic ranges and resolutions need parallel development.
(2) Production of mirror segments
This potentially major issue is already attracting serious study, as the sheer acreage of optical surface required is such that an efficient production system must commence work well in advance of the telescope's construction if its aperture is to be even partially filled at the time of commissioning. Thus considerable (and illuminating) discussion of production techniques is incoprporated in the proposal-paper for OWL (Gilmozzi et al. ,1998) which at that time was to include 2058 hexagonal spherically-curved segments, each 2 m across. Gilmozzi et al. focus on various fused silica derivatives and ceramics, and aluminium, as potential materials. Encouragingly, a new fabrication approach may make SiC (probably a variant comprising infiltrated Carbon/Silicon-carbide ``C-SiC'' technology) a serious option, which, if costs can be constrained, would offer huge reductions in the weight of the mirror segments and increases in their intrinsic stiffness. Ideas for the development of a ``production line'' process for glass-ceramics such as zerodur are alos presented. Fabrication by replication may also have the potential to achieve the high production rates required if the telescope is to be completed in a reasonable time. Variants of conventional polishing techniques, again aimed at a production-line approach whereby four segments are processed at once, were also investigated.
A discussion of the issues involved in fabricating aspheric segments for CELT is given by Mast, Nelson & Sommargren (2000). Considerations of cost and speed lead them to select a design incorporating ~1000 segments, each 50cm on a side (~1m across) and 45mm thick. Polishing would be done to a spherical target surface after the blank has been mechanically distorted so that release of stress after completion gives the desired hyperbolic form. Final ion figuring and fine-tuning using a Keck-style warping harness are proposed.
It should be noted that recent experience with other very large projects
indicates that, for spherical segments, mass manufacture may not
dominate production timescales: the ``Megajoule'' project acquired
~4000 sq m of good-quality spherical segments in ~3 years. However
this is most unlikely to be true of aspheric segments and this factor alone
may limit the size of telescopes with non-spherical primary mirrors.
(3) Multi-Conjugate Adaptive Optics (MCAO, AKA Atmospheric Tomography).
As indicated above, MCAO is likely to be the critical technology without which ELT projects may simply lack the attraction necessary to carry through to completion. That it can be made to work as proposed was demonstrated by Ragazzoni et al. (2000). In brief, this approach utilises several guide stars (GS: natural (N), or artificial, the latter most likely generated by exciting the atmospheric sodium layer at 90 km altitude with a suitably tuned laser(L)). From 4 to 9 suitably spaced N/LGSs are employed to eliminate the ``cone effect'' which limits the performance of classical AO using a single N/LGS, and to allow a similar number of wavefront sensors, controlling a smaller number of deformable mirrors conjugated to the altitudes of the main turbulent layers in the atmosphere, to sample, and determine the optical properties of, the entire volume of air through which the incoming wavefronts must pass. Technical issues, and solutions for VLT-class telescopes, are set out for example by Rigaut, Ellerbroek & Flicker (2000).
The MCAO technique consequently allows a much larger FOV to be corrected than is possible with a single GS. At least 3 natural GSs are nevertheless required to correct tip-tilt effects in low-order errors (e.g. anisotropic defocussing, astigmatism) (c.f. Rigaut et al. 2000). Because these are tip-tilt effects the isoplanatic angles* for the NGSs are large (in proportion to the telescope aperture) and this aspect of the overall design scales very benignly with increasing aperture. However earlier hopes of using only NGSs for correcting an ELT with wide sky coverage have been dashed (c.f. Rigaut et al. 2000, section 6). The dearth of natural guide stars probably implies that for adequate sky coverage by an ELT some use of laser GSs will be essential, unless schemes being developed by Rigazzoni and colleagues (Rigazzoni, Farinato & Marchetti, 2000b) prove successful.
* Isoplanatic angle: The angular separation of a guide star from the target object over which the correction derived from that star still improves the image from the target object.
At present it appears lthat at wavelengths >1 um Strehl ratios around 50% can in principle be achieved over fields of view ~1' , at least on 30- to 50-m telescopes. Strehl ratios of order 80% should be attainable at 2 um with this FOV; indeed at this wavelength Strehl ratios around 50% may be attainable over fields approaching 3' (B.Ellerbroek 2000, personal communication).
While the basic performance scales well with aperture there are technical issues of which this is not true. Because of perspective effects (seen from sub-apertures near the periphery of the entrance pupil the laser spot becomes a streak) the power requirements of the lasers for continuous LGSs scales as the square of the aperture, and probably become prohibitive fairly rapidly. If the use of LGSs is as necessary as suggested above, new techniques will be needed. The use of a pulsed laser can shorten the streak back to a moving spot, observable by gating the detector (perhaps optically, using a membrane mirror: F. Rigaut, 2001 personal communication). Saturation problems may constrain this approach, as the volume density of sodium atoms available for excitation is finite. New wavefront sensors, such as those discussed by Ragazzoni et al. (2000b) may help here as well as in the enhanced access to NGSs which these authors hope to achieve.
Scaling of the computer power required to manage the AO system, using
current methods, is expected to rise as the 4th power of the aperture,
and even Moore's law may not rescue the situation in time. However new
computing techniques
may
change the scaling law to DlogD, so it
may be that this technology can be matched to realistic requirements on
the required timescale. Nevertheless its attainability on a 100-m telescope
is quite a long way from credible demonstration. Again. Ragazzoni et
al. (2000b) hope to improve this situation by new sensing approaches.