New instrument concepts for observational cosmology

New Astronomy Reviews 50 (2006) 237–243 www.elsevier.com/locate/newastrev

New instrument concepts for observational cosmology J. Bland-Hawthorn Anglo-Australian Observatory, P.O. Box 296, Epping, NSW 2121, Australia Available online 17 April 2006

Abstract A key goal of observational cosmology today is to understand what initiated the re-ionization epoch and how this spectacular phase unfolded over a period of a few hundred million years. Simulations show that this may ultimately require us to push observations to redshifts as high as z  25. Here, I explore present and planned activities that will allow us to go beyond our current redshift limit (z  6.5). We stand at the dawn of a new era where diffraction-limited observing will be possible on 8 m class telescopes at near infrared wavelengths. I describe some of the instrument concepts that lead naturally from the science cases, in particular an AO-assisted, OHsuppressed IFU spectrograph. The benchmark for these new concepts has been set by the James Webb Space Telescope (JWST). Can we expect the next instruments to live up to this goal? If we do succeed, a great deal of entirely new science will be possible long before the expected launch of JWST.  2006 Elsevier B.V. All rights reserved. Keywords: Observational cosmology; Astronomical instrumentation; Photonics; Tunable filters

Contents 1. 2. 3.

4.

5.

6.

7.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Science goals – the big questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Technology goals – key instrument parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Sensitivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Adaptability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sensitivity through narrowband observation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Tunable filters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Tunable echelle imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sensitivity through OH suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Why is it so important? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. A new approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wide field advantage – implications for high spectroscopic and spatial resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. The traditional AX advantage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. The maximum AX product in a narrow spectral band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3. Adaptive optics: a new direction in AX advantage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

E-mail addresses: [email protected], [email protected]. 1387-6473/$ - see front matter  2006 Elsevier B.V. All rights reserved. doi:10.1016/j.newar.2006.02.019

. . . . . . . . . . . . . . . . . . .

238 238 238 238 239 239 239 239 239 239 240 240 241 241 241 242 242 243 243

238

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

1. Introduction

2. Science goals – the big questions

The structure of this meeting has emphasized the close bond of theory, experiment and observation. This has always been true for the frontier of observational cosmology which, more than any other astrophysical pursuit, has been the driving force for bigger telescopes and better machines over the past 50 years. The microwave background experiments (COBE, Boomerang, WMAP, VSA) have shaped the observational frontier in two distinct ways. First, they have detected the signature of primordial structure and subsequently determined accurate measurements of the Standard Model parameters. This has allowed numerical simulators to concentrate on the more difficult gas physics rather than be left with cosmological uncertainties on the basic framework. Secondly, the WMAP polarization measurements suggest a possible reionization epoch as early as z  20, although this value may well come down in subsequent reanalysis (cf. MacTavish et al., 2005). The most recent simulations suggest the first bound structures formed at z = 26.0, the first star at z = 25.2, with the first light and first HII regions at z = 25.0 and z = 24.7, respectively (Yoshida et al., 2004). We now talk in terms of First Light. This has provided the impetus to numerous experiments from gamma ray to radio wavelengths given that the Dark Ages might not be so dark after all. At optical and infrared wavelengths, we do not need to await the James Webb Space Telescope (JWST) to make substantial progress in this arena. Existing 8 m telescopes can and will be used more effectively to pursue the Dark Ages, even before the launch of the JWST. Diffractionlimited performance is close at hand, and should come close to delivering D4 performance (where D is the telescope diameter) for at least some of the planned instruments. The focus of this meeting has been on 3D integral field spectroscopy. It is remarkable how far this technology has come since the first 3D spectrograph conference in Marseille (Bland-Hawthorn, 1995). This technology has matured considerably over the past decade and is widely employed on front line instruments. Here, we show that the combined impact of adaptive optics and new developments in photonics is expected to greatly enhance the sensitivity of near-infrared integral field spectrographs. The structure of this paper is as follows. In Section 2, we summarise the key goals of First Light science. These lead directly into the technology goals listed in Section 3. In Section 4, we discuss the importance of tunable imaging filters. In Section 5, we explore sensitivity improvements in future integral field spectrographs through photonic OH suppression. In Section 6, we revisit the traditional meaning of wide-field (AX) advantage, and describe implications for high resolution work.

There are many things that we would like to know about the Dark Ages. What initiated the epoch of reionization and how did it develop? Did it happen in stages due to more than one type of source? Are the halo ultra metal poor stars we see today the byproduct of the First Stars and therefore reveal the metal products of this unique population? It is widely believed that the bulk of the galaxy spheroid population formed around z  3 and the disks at z  1, but unequivocal evidence for this scenario in the far field is still not secure. Can we demonstrate this more directly by establishing the progenitors of both systems? How did galaxies assemble before this time and where does the present day bulge – black hole mass relation come from (Magorrian et al., 1998)? It is important to keep in mind that observational cosmology at these redshifts has the potential to provide insight on the nature of dark matter and dark energy, two of the great unknowns of contemporary astrophysics. Yoshida et al. (2003) illustrate how the evolution of structure in the early universe can indicate the presence of warm dark matter. At lower redshift, oscillations in the power spectra of future large-scale redshift surveys may shed light on the nature of dark energy (Seo and Eisenstein, 2003; Blake and Glazebrook, 2003). 3. Technology goals – key instrument parameters For ground-based astronomy, there has been a great deal of technological progress in the past decade: 8 m telescopes, large format high efficiency CCDs and IR arrays, multi-object and integral field spectrographs (the subject of this meeting), high performance dispersers (e.g. tunable filters, volume phase holographic gratings) and differential techniques (e.g. iodine cells, nod & shuffle). When we look at the numerous scientific discoveries over the same period, we can identify four key factors which have made these possible: sensitivity, resolution, efficiency, and adaptability. These factors are fundamental to future scientific progress. 3.1. Sensitivity During the so-called Dark Ages, several factors conspire to make objects darker: (i) distance and cosmological dimming, (ii) pervasive neutral medium scattering resonant emission lines, (iii) dust. In order to achieve the First Light science goals, ground-based telescopes will need to achieve much higher levels of sensitivity than have been possible before. Traditionally, this has meant pushing for higher instrument throughput and larger telescope apertures. Modern instruments often have total system efficiencies of 30% or higher; the only substantial instrument gains to be had are for niche applications (e.g. Baldry et al., 2004). Since the next telescope generation is a decade away, the greatest sensitivity gains in the short term must come from suppressing the night sky emission and diffraction-

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

limited observation. Since most deep observations (involving imaging or low to moderate resolution spectroscopy) are sky background limited, this remains the fundamental limit across the R, I, Z, J, H and Ks bands. A device which is able to suppress the background with high efficiency at J and H will have major consequences for First Light science; a device which can do this with diffraction-limited performance will even rival the JWST. 3.2. Resolution A parallel approach to increased sensitivity from the ground is diffraction-limited observation assisted by adaptive optics. This enhances the contrast between the observed source and the background and is being implemented on all 8 m telescopes, and will be a cornerstone technology of extremely large telescopes. Instrument concepts properly optimised for AO-corrected light are a relatively new innovation. Higher angular resolution also provides higher intrinsic spatial resolution at a fixed redshift. 3.3. Efficiency The most spectacular efficiency improvements in recent years have come from the huge multiplex advantage offered by wide field, multi-object spectrographs (e.g. 2dF, 6dF, SDSS). The same holds true for First Light science. In particular, the narrow band and photometric studies reveal many candidates which need to be followed up with a multi-object spectrograph and very long exposures. 3.4. Adaptability A parallel approach to increased efficiency is adaptive sampling (or tiling) of the focal plane. Not all science cases are well served by a contiguous field, or by hundreds of single element probes which typically cannot be clustered to a few locations in the focal plane. For example, in terms of resolved spatial elements, only about 10% or less of the pixels in the Hubble Ultra Deep Field (HUDF) contain useful information. Independent of the multiplex advantage, a device that can adaptively sample the focal plane can greatly improve the efficiency of First Light science programmes. This topic is discussed in some detail in BlandHawthorn et al. (2004).

239

larly those which are directed at fields with pre-existing, broadband data. But a more sensitive and versatile approach to narrowband imaging than the use of conventional filters is provided by a tunable imaging filter. This was amply demonstrated over a 7 year period by the Taurus Tunable Filter (TTF; Bland-Hawthorn and Jones, 1998) at the Anglo-Australian Telescope and William Herschel Telescope. The results from these programs have been reviewed by Bland-Hawthorn and Kedziora-Chudczer (2003). The TTF allowed accurate differential measurements between neighbouring spectral bands, primarily because both the optical path and the spectral response of the interference band is essentially identical everywhere over the free spectral range (roughly the bandpass of the order selecting blocking filter). This was further aided by charge-shuffling and nod & shuffle techniques. Francis and Bland-Hawthorn (2004) show the extraordinary depths you can reach, even on a 4 m telescope in moderate seeing. Optical devices are under development for the Magellan telescope, the South African Large Telescope, and the GranTeCan telescope (Cepa et al., 2003); a tunable filter is also under development for the NTT telescope. All of these facilities will target the high redshift universe. Fabry–Perot devices have been used at near-infrared wavelengths for Galactic and extragalactic work (e.g. Wright et al., 1989). For a ground-based device, the many hundreds of bright atmospheric OH lines are a challenge to a tunable filter with its characteristic Lorentzian wings (high finesse limit). This contrasts with an interference filter where the wings can be damped by the use of multiple resonant cavities (Jones et al., 1996). The wings of a tunable filter response function can also be damped by cavities in series, and such possibilities are under discussion (R. Abraham 2004, personal communication). Contrary to what is sometimes stated at meetings, there is no strong scientific case for an infrared imaging Michelson operating from the ground, at least for extragalactic work. All OH lines contribute noise at all frequencies in Fourier space, such that all spectral channels are now affected once the data are inverse-transformed back to wavelength space. Above the Earth’s atmosphere, the case for an imaging Michelson or any other tunable filter technology is more compelling (for a review of technologies, see Bland-Hawthorn, 2000). Indeed, an infrared tunable filter is under development for the James Webb Space Telescope (Satyapal et al., 2000) with the specific aim of searching for high redshift Lya emitters.

4. Sensitivity through narrowband observation 4.2. Tunable echelle imaging 4.1. Tunable filters Cosmological surveys of emission-line sources (e.g. Ouchi et al., 2005; Hu et al., 2004; Kudritski et al., 2000; Jones and Bland-Hawthorn, 2001; Baker et al., 2001; Barr et al., 2004; Glazebrook et al., 2004) underline the great potential of targeting narrow photometric bands, particu-

There are interesting hybrid techniques which should be explored at a time when infrared detector real estate is considered relatively inexpensive in the context of the overall cost of an instrument. One of these is called ‘tunable echelle imaging’ (Baldry and Bland-Hawthorn, 2000) where the output from an imaging Fabry–Perot interferometer is

240

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

cross-dispersed by a grism in one direction, and dispersed by an echelle in the other direction. This technique forms a mosaic of different narrowband images of the same field on a detector. The technique does not pack the output at the detector in a particularly efficient manner, and is therefore moderately wasteful in pixels. There are also important considerations about the psf which can be moderately elliptic. However, those images which are compromised by OH sky lines can be simply ignored. 5. Sensitivity through OH suppression 5.1. Why is it so important? At the present time, broadband sensitivity at near infrared wavelengths is much poorer than at optical wavelengths. In an 8 h exposure, the deepest galaxy surveys to date reach to only 24 mag (5r) in the near infrared compared to 28 mag in the optical. The highest redshift sources (z 6.6) are associated with where the Lya line falls at the limit of optical detector sensitivity. To venture beyond here into the Dark Ages (z > 7) with the current generation of 8 m telescopes demands that we achieve optical levels of sensitivity in a broad band at near-infrared wavelengths,

1.5

1.52

something which has not been achieved to date. In the following section, we highlight a promising new approach to OH suppression which makes use of recent advances in photonics (see Fig. 1). If the ‘‘gold standard’’ is set by the expected performance of the HAWAII- 2RG arrays under development for the JWST, then we arrive at an important result for Dark Age science. As we show in Section 5.2, OH suppression is demanded for resolutions of R  5000, well into the regime commonly associated with software suppression techniques. In any event, the required spectral resolution is substantially lower. At redshifts of z = 7–10, the Lya line width is expected to be 100–200 km s1, depending on whether the plasma is photoionized in a starburst, arises in a wind shock or from ram pressure stripping in a merger. Thus, in the first phase of detecting these systems, an optimal resolution is R  1000, although follow up studies will need R  3000. These are the expected resolutions of the NIRSPEC infrared spectrograph on the JWST. However, from the ground, both resolutions demand full OH suppression. Let us consider a million element integral field unit (MEIFU), a challenging device first considered by Content et al. (2003). It is worth keeping in mind that such devices,

1.54

1.56

0 -10 -20 -30

4000 2000 0

6000

4000

2000

0 1

1.2

1.4

1.6

1.8

2

Fig. 1. The top panel is the wavelength response of a new technology OH-suppressing fibre developed and patented by the AAO and industrial partner Redfern optical components. The middle panel shows how the rejection bands line up perfectly with the H-band sky spectrum. The bottom panel illustrates the region of the spectrum treated by our first prototype (BEE).

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

albeit of a more rudimentary form, already exist at the target plane of particle physics accelerators. An IFU spectrograph with photonic OH suppression exploiting, say, a pixel format of 1024 · 1024 fibres does not need to disperse at all to have fundamental application. For example, if we image all of the light in the J or H band without dispersing, and allow 4 · 4 pixels for each fibre output, this requires a single 4K · 4K array. If OH-suppressed broadband imaging could be achieved with diffraction-limited performance, such a device would rival the performance of the JWST. We can also consider a MEIFU with moderate spectral dispersion, if we specify a resolution of R = 1500 for an H band spectrometer (say 1350–1850 nm) with two pixels per spectral resolution element, and two pixels per spatial element assuming that the neighbouring fibres on the sky can be packed this closely on the detector. For an IFU format of 256 · 256 fibres, the total detector requirement is eight 4K · 4K arrays. 5.2. A new approach Different approaches have been taken to OH suppression. These include spectrographs with notched gratings, and interference filters with a notched wave-length response. However, the demonstrated or expected gains are not likely to achieve the First Light science goals. (For the case of OH suppression spectrographs, the overall throughput is less than 5% in all cases). For the most part, these methods are OH avoidance schemes rather than suppression schemes. The OH lines have intrinsic widths that resolve out at R  150, 000 and so need to be suppressed at resolutions where the amount of available filter band removed is negligible, i.e. R  10, 000 or greater (Jones et al., 1996). We have been exploring a new technology based on the fibre Bragg grating (FBG; see Fig. 1). This has now been demonstrated in the laboratory in single mode (BlandHawthorn et al., 2004 – BEE) and multi-mode fibres

241

(Bland-Hawthorn, 2005). The single mode fibres are applicable to diffraction-limited performance, and the more complex multimode fibres are for lower Stehl ratios and natural seeing. Integral field spectrograph concepts which exploit these new fibres are now being explored. In Fig. 2, we demonstrate the importance of OH suppression. The lefthand figure shows the wing behaviour in arc lines taken with the IRIS2 infrared spectrometer at the AAT. This Lorentzian behaviour can be understood in terms of classical light scattering in grating spectrographs. If we then convolve this psf with the OH spectrum (righthand figure), we can see that there are few spectral channels clear of OH emission even at R = 5000. The dashed line is the expected dark current of the HAWAII2RG arrays under development for the JWST. This is the ‘‘gold standard’’ by which sensitivity must be gauged in an era of Dark Age astronomy. 6. Wide field advantage – implications for high spectroscopic and spatial resolution 6.1. The traditional AX advantage Survey efficiency is often loosely stated in terms of the AX product, where A is the telescope area, and X is the total solid angle of the sky survey. For a single telescope pointing,  2 2 pd A1 X1  A2 X2  ð1Þ 4F 2 where d is the detector size in mm, and F is the beam focal ratio (0.5 6 F < 1). The subscripts are defined in Fig. 3. For AX to be a useful measure of survey efficiency, there are certain qualifications. We assume that the sources under study are: (a) resolved by the instrument and the pixel sampling, (b) detected in a reasonable exposure time, and (c) the signal-to-noise ratio (SNR) 1 increases as ðtimeÞ2 .

Fig. 2. Left: near infrared observation of Xe arc lamp taken with IRIS2. The log intensity scale emphasizes the Lorentzian wings on the brightest arc lines. The effective resolving power is R = 2400 so that the arc lines are almost optimally resolved. Note how the faint wings extend over more than 20 pixels. The inset shows our model for the scattered wings (Lorentzian component is 8% by total flux). Right: logged H-band sky spectrum convolved with IRIS2 scattering model (left) at two different resolutions (R = 1500, 5000). The dashed line is the expected dark current (1 h exposure) in the HAWAII-2RG arrays under development for JWST. The curves below the dashed line illustrate the impact of OH suppression.

242

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

Fig. 3. Left: Properly matched optical system where the A1X1 product of the detector measured at the telescope aperture is equal to the A2X2 product of the telescope measured at the detector. The telescope diameter is D, and the focal length of a spherical mirror is f; the detector size is d. The AX invariant is the throughput or e´tendue of the system. Right: Simulation of the on-axis response of a 1%k filter in different converging beams. The centroids and dispersions are also shown as derived from theory (Bland-Hawthorn et al., 2001). The inset figure shows the degradation of the filter response at an off-axis position in an f/2.5 beam.

For any telescope, fast beams are much more efficient for survey work. In Eq. (1), to sidestep subtleties involving field-expanded foci, we adopt F P 2 as faster beams cannot be properly compensated by Lyot’s method. But fast beams degrade the energy interval isolated through spectroscopic interference, particularly for interference filters. Interestingly, this is not aided by going to collimated beams since the focal reducer preserves the field angle h of the incoming beam. Since the path length has an angular dependence, interference filters exhibit a phase effect (shifting central wavelength) over the field of view. Traditionally, wide-field surveys have been restricted to broad photometric bands. Substantially, narrower bands provide a technical challenge. In the context of First Light science, what constitutes a wide field? Recent Subaru observations with SuprimeCam have revealed that Lya sources at z  6 show clear large-scale structure over the degree-sized field of view (Ouchi et al., 2005). These results were established in the 815 nm optical window with a ˚ filter in an f/2 beam. 120 A The detection sensitivity could be at least a magnitude deeper if it was possible to place a substantially narrower band in the same beam. However, in Fig. 3, we show the degradation of a 1%k filter in the Subaru f/2 telecentric beam. The response is visibly poor for this filter, and would be ten times worse for a 0.1%k filter (peak transmission approx 10% for a filter with close to 100% transmission in a collimated beam) like those used in the DAZLE imaging spectrograph (Horton et al., 2004).

ously degraded in fast beams. Prima facie, this appears to argue against tunable filters as efficient cosmological surveyors. So is it possible to retain the high performance of a narrowband filter in an f/2 beam? Remarkably, the answer is a resounding ‘‘yes’’ and for an arbitrarily narrow bandpass. Lyot (1944) showed that beams as fast as f/2 could be compensated by crossed birefringent elements such that even sub-Angstrom, wide-field images are possible. This forms the basis of the beam-correcting filter design presented here. Based on this principle, Bland-Hawthorn et al. (2001) describe a tunable filter for use in an f/3 beam (AAT prime focus) which makes use of polarizing (P) and retarding elements (R) which rotate differentially. But we can envisage a much simpler monolithic Lyot¨ hman filter (non-tunable) which would comprise an O ˚ blocker (B), and a series of polaroid and retarding 80 A layers in an arrangement B:P:R:P:2R:P:4R:P:8R:P. Some of the fattest layers would need to be split and arranged as shown in our paper (Bland-Hawthorn et al., 2001, Fig. 6). So the resulting filter would require between 10 and 18 thin layers of birefringent material. The polishing of these layers, even over the required 200–300 mm aperture, is mildly challenging and within reach of a precision optics group. The main concern is to obtain birefringent material of sufficient quality and low birefringence ˚ filter without (0.01). The end product would be a 10 A degradation or phase change over the entire 1 field. This would be a moderately high risk albeit expensive undertaking.

6.2. The maximum AX product in a narrow spectral band 6.3. Adaptive optics: a new direction in AX advantage As we have seen, the ‘‘area–solid angle’’ (AX) product is a useful measure of survey efficiency. This quantity depends inversely on the square of the telescope f/ratio which is unfortunate since narrow spectral bands are seri-

Traditionally we speak of ‘‘wide field’’ in terms of the largest possible angle that can be achieved at a given telescope focus. However, in an era of adaptive optics, one

J. Bland-Hawthorn / New Astronomy Reviews 50 (2006) 237–243

should embrace another interpretation of ‘‘wide field’’ which is measured in pixels rather than in solid angle. When one talks of wide-field AO, we mean the largest possible pixel format which properly samples the corrected psf, regardless of the pixel sampling. For AO-corrected imaging, it is useful to consider different levels of correction, i.e. low order correction (e.g. ground layer AO) and high order correction, e.g. multi-conjugate AO). In the first case, we can expect 0.200 corrected seeing, whereas the latter should approach the diffraction limit of the telescope. Lyman Break Galaxies at z  3 3 demonstrate that clustering scales are 5–10 Mpc so to fully characterize this structure requires that surveys extend over twice this scale. Thus, deep survey fields in 0.200 (200 mas) seeing should be of order 10 0 at z = 1 down to 4 0 at z = 6. This leads to a fully sampled pixel format of about 6000 · 6000 pixels. For the Dark Ages (z > 7), spatially resolved studies in deep survey fields will require 2 0 fields fully sampled at 10 mas or better leading to a pixel format of 12,000 · 12,000 pixels. It is noteworthy that the pixel format tends to increase as the number of corrected modes is increased. For example, an extreme AO system achieving 1 mas sampling can reasonably expect to reach 2000 or a format of 20,000 · 20,000 pixels. With some effort, science cases can be made for the different levels of AO correction, but these are not summarised here due to lack of space. It pays to consider the science case in the context of an instrument concept (imaging or spectroscopy, contiguous or multi-object) and to understand how the observations will be compromised if the AO correction falls short of its design specification, which it may well do. In passing, it is worth noting that the case or OH suppression becomes progressively weaker as we go to increasingly higher order AO correction. In this respect, one can think of adaptive optics as providing us with a form of sky suppression through extreme magnification. However, fully diffraction-limited AO correction is still some years away, particularly in the J and H bands. This instrument concepts must consider the implications of near diffraction-limited performance rather than the idealized case. 7. Conclusions Here, we have considered the main technology goals behind scientific advance: sensitivity, resolution, efficiency and adaptability. Improvements in one parameter often aid the others. As we have shown, there is scope for major improvements in some of these parameters, in particular, sensitivity through total suppression at R = 10,000 of the atmospheric OH lines.

243

The next generation of extremely large telescopes will be heavily dependent on very sophisticated forms of adaptive optics correction. (It is interesting to note that the James Webb Space Telescope will also require some form of AO correction.) It will be very important for these systems to work close to their design goals in order compete or surpass the performance of the JWST in the next decade. Arguably, the development of instruments properly adapted to AO-assisted observation is in its early stages. It seems plausible that the ongoing and future delays in JWST and ELTs will lead to 3rd generation instruments for the current 8 m telescopes. Therefore, it would seem wise that we consider new instrument concepts which fully exploit low and high order AO correction, and that we consider ‘‘killer applications’’ for these new concepts. We have suggested a few avenues of research here, with the most promising being photonic suppression of atmospheric OH lines in future integral field spectrographs. References Baker, J.C. et al., 2001. AJ 121, 1821. Baldry, I.K. et al., 2004. PASP 116, 403. Baldry, I.K., Bland-Hawthorn, J., 2000. PASP 112, 1112. Barr, J.M. et al., 2004. AJ 128, 2660. Blake, C., Glazebrook, K., 2003. ApJ 594, 665. Bland-Hawthorn, J., 1995. In: 3D Optical Spectroscopic Methods in Astrophysics, ASPC, vol. 71, p. 369. Bland-Hawthorn, J., 2000. In: Imaging the Universe in 3D, ASPC, vol. 195, p. 34. Bland-Hawthorn, J. et al., 2001. ApJ 563, 611. Bland-Hawthorn, J. et al., 2004. SPIE 5492, 242. Bland-Hawthorn, J., 2005. AAO Newsletter 108, 4. Bland-Hawthorn, J., Jones, D.H., 1998. SPIE 3355, 855. Bland-Hawthorn, J., Kedziora-Chudczer, L., 2003. PASA 20, 242. Bland-Hawthorn, J., Englund, M.A., Edvell, G., 2004. Optics Express 12, 5902. Cepa, J. et al., 2003. SPIE 4841, 1739. Content, R. et al., 2003. SPIE 4842, 174. Francis, P.J., Bland-Hawthorn, J., 2004. ApJ 353, 301. Glazebrook, K. et al., 2004. ApJ 128, 2652. Horton, A.J. et al., 2004. SPIE 5492, 1022. Hu, E. et al., 2004. AJ 127, 563. Jones, D.H. et al., 1996. PASP 108, 929. Jones, D.H., Bland-Hawthorn, J., 2001. ApJ 550, 593. Kudritski, R.-P. et al., 2000. ApJ 536, 19. Lyot, B., 1944. Annales D Astrophysique 7, 31. MacTavish, C.J., et al., 2005. Available from: astro-ph/0507503.. Magorrian, J. et al., 1998. AJ 115, 2285. Ouchi, M., et al., 2005. Available from: astro-ph/0508083.. Satyapal, S. et al., 2000. ASPC 207, 212. Seo, H.-J., Eisenstein, D.J., 2003. ApJ 598, 720. Wright, G.S. et al., 1989. In: Infrared Spectroscopy in Astronomy. ESA pub., p. 425. Yoshida, N. et al., 2003. ApJ 591, L1. Yoshida, N. et al., 2004. ApJ 605, 579.