The programme proposed by Mountain (1996) for initial definition of the science goals of an ELT is the need to obtain spectroscopy of all the galaxies in the Hubble Deep Fields. This can be achieved by a 50-m telescope operating near the diffraction limit over armin FOVs: indeed, by using Integral Field techniques spectra of the components of such galaxies should be obtainable. However, objects which are more thinly distributed on the sky could in principle justify a much wider field, to be covered at the seeing limit by a multi-object spectroscopic system.
MCAO may offer Strehl ratios > 50% over ~arcmin fields at wavelengths around 2 um, perhaps 30 times higher than for a seeing-limited image. The speed with which the telescope can secure a given observation of a faint source is proportional to the square of this ratio: the ELT should therefore be almost a thousand times faster for faint-object spectroscopy in MCAO mode than in seeing-limited mode. Thus 100 widely-scattered targets would be observable several times faster in MCAO mode than in seeing-limited mode with a multi-object instrument, even allowing acquisition overheads. This result is strongly aperture-dependent. Barden and Parks (1999) derive a relative speed advantage for an MCAO-equipped 30-m ELT of only 100 relative to the seeing-limited case, so overheads would make serial observation of 100 objects significantly slower with the MCAO than their (parallel) observation with a MOS system. This is because speed goes as the 4th power of aperture under these (approximately diffraction-limited) observing conditions, giving a 50-m an 8-fold speed advantage over the 30-m.
However all else is by no means equal: the seeing-limited facility must
fit several hundredths of a square arcsec onto a typical detector pixel.
For a 50-m telescope to squeeze a 0."2 angular FOV onto a 20 um-square
pixel a final f-ratio of 0.41 is required (c.f. Atad-Ettedgui et
al., 2000, section 6.1.1). It is generally considered that designs
having final f-ratios < 1 are unlikely to be realisable. Substantial
oversampling of the delivered images will necessarily have to be accepted:
a 50-m with final f/ratio of 1 would have ~6 pixels across the image FWHM,
resulting in considerable performance losses for spectroscopy. Currently
it is difficult to see how a 50-m class ELT can be used efficiently in
a seeing-limited mode, unless this is viewed as a backup mode in which
loss of performance can be accepted in order to stay in operation. Nevertheless
this debate is quite unresolved, and will probably remain so until a clear
and hard-nosed set of scientific objectives (c.f. the ``Design Reference
Mission" of the NGST) are examined in detail.