Chapter 2 summarizes the frontier science enabled by the enormous gains in sensitivity and angular resolution afforded by a 30-m class telescope. During the next year, the New Initiatives Office (NIO) will work with the community to further explore the potential of GSMT, to quantify the gains of GSMT compared with extant and planned facilities, and to understand more fully the key roles of a GSMT in the era of James Webb Space Telescope (JWST) and Atacama Large Millimeter Array (ALMA). To inform our point design choices, we adopt the science cases described in Chapter 2 as an initial guide to defining the key performance characteristics of our point design.
Meanwhile, in parallel, we are developing the GSMT Point Design telescope driven by the science described above. Later in this Book, we will discuss the philosophy behind the particular design choices made in the point design. In this section, we discuss the specific requirements that the science places on the design.
We also assume a 30-m filled-aperture telescope as a starting point: it is large enough to enable qualitatively new science (see Chapter 2), challenging technically, with cost and risks within the bounds of extrapolation based on experience, and possible to complete contemporaneously with ALMA and JWST.
The science opportunities described in Chapter 2 are:
- Large scale structure
- Galaxy formation
- Resolved stellar populations
- Planet formation environments
- High dynamic range science
Table 1 abstracts the observing requirements on a 30-m GSMT for the programs described.
Key requirements extracted from Table 1 include:
- Diffraction-limited (DL) performance from the thermal IR down to the shortest feasible wavelengths, for imaging and spectroscopy, over moderate fields
- Spectroscopic capability in the mid-IR (at very high resolution)
- Multi-object, wide-field performance at optical wavelengths
These are specifically areas of performance where the GSMT will be complementary to capabilities of the JWST: The GSMT will have wide-field (20') optical wavelength capability; its spatial resolution will be set by the diffraction limit of its 30-m aperture; and it will be able to accommodate large, high-spectral-resolution spectrometers.
The diverse requirements for the different science opportunities will be addressed by using the telescope in different modes as defined in Table 2. Reference to more detailed discussion of these modes is appended in each case. Here we confine ourselves to enumerating the observing modes and citing references to further discussion later in the text.
|Table 2 Observing modes defined.|
|Prime focus AO||Visible light, seeing-limited, fiber-fed spectroscopy. Here the AO is for control of the primary figure and correction of ground-layer turbulence only. (Appendix 4.6.A)|
|Direct Cassegrain focus AO||2-mirror, low-emissivity, high-throughput, relatively narrow field. This mode is well-suited for near- to mid-IR narrow field spectroscopy. (Section 22.214.171.124)|
|Narrow Field Very High Order AO||Also at Cassegrain focus, with the addition of another very high-order adaptive element, this mode enables high Strehl, near- to mid-IR, high dynamic-range imaging and spectroscopy over a narrow field of view. (Section 126.96.36.199)|
|Multi-conjugate AO||With the addition of several laser guide stars and additional multiple adaptive mirrors, this mode enables near-IR imaging and multiple object spectroscopy over moderate field of view. (Section 4.6.2)|
The list of modes is not exhaustive, but spans the possibilities that have been developed in any detail so far. Another possibility, for example, is to develop the direct Cassegrain over the 5-arcmin field available without vignetting for a seeing-limited or partially AO-compensated mode.
The implementation of these modes in the point design will be presented in Section 4.1. A suite of instruments has been designed (see Section 4.7) to exploit the strengths of the telescope in its different modes and to assess: (1) the challenge of accommodating the instruments to the telescope, and (2) the feasibility of designing instruments for the Point Design telescope
Prime focus AO - The driver for this mode is to exploit the enormous niche that is provided by the visible spectrum and the circumstances when DL correction is not possible (thin cirrus, unfavorable turbulence profile above the site, etc.). Thus, the requirement on image quality is relaxed to include seeing-limited, as well as adaptive "enhanced seeing" conditions. Given the mismatch in plate scale at the Cassegrain focus to plausible visible light detectors under seeing-limited conditions, we propose implementing this mode at prime focus. To implement an "enhanced seeing" mode, wherein low altitude turbulence and wind-buffeting are selectively and partially corrected, several natural guide stars (NGSs) will be used for wavefront sensing. Because of the desire for relatively high resolution spectroscopy, the spectrograph(s) for this mode will be located remote from the focus, thus requiring fiber transport of the light from the focal plane to the spectrographs.
Direct Cassegrain focus AO - Implementation of high-resolution thermal-IR spectroscopy. The requirement here shifts to low thermal background and high throughput. Our approach makes use of an adaptive secondary to produce correction to fairly high order, so with an NGS and a Wavefront Sensor (WFS), excellent light concentration (high Strehl) will be available to feed a high-resolution thermal-IR spectrograph. Note that in the thermal IR, the use of an adaptive secondary permits achieving high Strehl without additional thermal emission.
Narrow-field high-order AO - Implementation of coronagraphic imaging. With the addition of a second deformable mirror (DM) of very high order, high Strehl performance can be extended to the near IR. The aim here is to maximize light concentration and minimize scattered light for high dynamic range imaging.
Multi-conjugate AO - This mode is designed to produce DL images over a field of view (FOV) of 1-2 arcmin. A large number of sizeable optical surfaces are required in order to implement this concept. In principle, this mode would be useful from the near IR to the shortest feasible wavelengths; in practice, it will be limited to the near IR because of technical difficulties in achieving significant Strehl ratios in the visible.
The requirements of FOV, wavelength regime, and spectral resolution are set by fundamental characteristics of the telescope and instruments. We will return to the matter of how these are met in the context of the telescope and instrument design. Spatial resolution is a product of telescope scale, pixel size, and most importantly, image quality or wavefront control.
The ability of the GSMT to deliver DL images to the instruments is its greatest challenge and its greatest strength. The need to match its increased light grasp with enhanced angular resolution is given urgency by the greater confusion of the fainter objects it will be observing. Achieving this performance represents a far greater challenge than for previous telescopes because the same or better level of wavefront control over the atmosphere and optical surfaces must be maintained over a much larger optical aperture. For given wind, thermal gradients, and gravity, their respective effects on distortion of the mirrors will be greater for a 30-m telescope than for an 8-10-m telescope by factors which depend at least linearly on the radius. Fortunately, despite the smaller ultimate DL image for the 30-m, the requirement on RMS (root-mean-square) wavefront error is independent of diameter. The Strehl ratio, in the near-diffraction-limited regime, is given approximately by the formula:
where OPD is the RMS wavefront error or Optical Path Difference (OPD), and is the wavelength. The Strehl ratio is the ratio of the intensity of the image of a point source delivered by the telescope to that of a perfect, DL image formed with the same telescope aperture and throughput.
|Wavefront error for Strehl of 0.3|
|Table 4 Wavefront error and Strehl|
However, the RMS OPD due to turbulence (refractive index inhomogeneities) grows with the diameter of the mirror. Neglecting important outer-scale effects, the OPD would grow as D5/3. The contribution to the error of the mirror figure will also increase because of the decreased stiffness (lower resonant frequencies) of the telescope structure and mirror support. The challenge of meeting the OPD requirement will be significantly more difficult than for telescopes of the current generation.
Sorting out the contributions to the wavefront error and designing a system that can keep it under control are fundamental to the concept of the GSMT. These topics will be discussed in detail in Section 4.2.
What are the requirements on image quality that determine the scientific productivity of the telescope? Table 5 discusses the principal areas of concern:
It is important to recall that although Strehl ratio measures light concentration in the image, it says nothing about throughput, which is also very important in determining the overall signal-to-noise ratio of an observation. AO systems inevitably require extra optical surfaces, and the enhancement in the shape of the PSF comes at some cost in its amplitude. In evaluating the delivered signal-to-noise ratio, it is important to include both effects. The requirements on throughput are discussed separately, below.
The last three characteristics of the PSF are important for accurate photometry. If the PSF is uniform, or smoothly varying according to a stable function of field position, one can reliably determine the PSF from bright calibrator stars and apply it over the field. If the PSF is stable, frame-to-frame, the PSF may be used without modification over a set of frames. Unfortunately, the PSF is inevitably affected by external disturbances-seeing or wind-buffeting, which tend to be extremely variable-so PSF stability is an ideal that can never be fully achieved. Because adaptive correction is always subject to a residual error that increases when the natural seeing or wind-buffeting is worse, the PSF is inherently unstable at a level determined by the amplitude of the external disturbances. Clearly, reducing the external seeing and stiffening the telescope against wind-buffeting will always be the first line of defense.
Another aspect of PSF stability that is serious for photometry in crowded fields or in high dynamic range imaging is speckle noise. Speckle is an aspect of the PSF that becomes apparent in near-DL conditions. Speckle is an interference pattern associated with low-level aberrations in the optical system, including the atmosphere. Although the atmospheric aberrations vary rapidly and the speckle pattern due to them will average out in a few seconds, those due to aberrations in the telescope may vary relatively slowly, leaving a residual speckle pattern in the images that will not be identical frame-to-frame. This is speckle noise. The phenomenon has been discussed by Racine.1 The ability to improve s/n in a high contrast scene via long integration rests crucially on both the magnitude of speckle noise and how contributions from this noise source "integrate down." This has been identified as an area for further study and is discussed and illustrated in Section 188.8.131.52 on high dynamic range imaging.
By predictability, we imply either of two concepts: (1) that the dependence of PSF on field position is well-behaved and can be interpolated based on the determination of the PSF at a few points in the field from stellar measurements, or (2) that measurements of the AO system (commands sent to corrective mirrors, for example) can be used to anticipate the quality (Strehl) of the PSF. The latter type of predictability would be useful and has been proposed and studied2,3 and implemented in PUEO on the CFHT and NAOS on the VLT, for example. It deserves further study, but we will not try to specify a performance goal at this time.
Setting requirements for the other aspects of image quality (see Table 6) requires tradeoffs in cost and scientific productivity that are difficult to make at this time. However, it is useful to state some goals for these aspects of performance that should permit a high level of scientific productivity. These goals are, however, dependent on the observing mode of the telescope. For each mode, we state the image quality requirements that we believe strike a balance between scientific payoff and engineering cost.
To achieve the science goals, the GSMT will normally be operated in a closed loop with sensors detecting the delivered wavefront directly. But to operate efficiently, the telescope must deliver basic performance at a level sufficient to fall within the capture range of the systems of control. Furthermore, the drive systems must operate smoothly; the temporal spectrum of residual errors and vibrations delivered to the mount by the drives will have to roll off to negligible levels at frequencies above about 1 Hz.
Horizon limit for pointing - 20 degrees
Pointing - The telescope mount and drives must be capable of pointing using encoders only to any point on the sky 20 degrees above the horizon to within 3 arcsec RMS, in the absence of wind loading.
Tracking - The telescope mount and drives should be able to track a point on the celestial sphere, open loop, using encoders only, in a manner which is consistent with the pointing requirement. In addition, the RMS amplitude of the tracking errors (in the absence of wind loading) in the frequency range above 1 Hz should be limited to less than 0.1 arcsec, which ensures that the drive errors will be smaller than but comparable to the wind driven pointing error. This is to minimize the demand on the adaptive control systems.
Active optics - The active optics will rely on signals both from the segment edge sensors and one or more coarse wavefront sensors in order to achieve the optical correction required to fall within the capture range of the fine WFSs, which will ultimately control the figure of the primary and secondary. We set that capture level to correspond to an RMS image size of 1 arcsec. Operational efficiency requires two other aspects of performance:
- Active optics update time - After the acquisition of a calibration star and an exposure, the time required for the coarse WFS to compute the corrections of the figure, and for the actuators to receive and respond to the correction commands, should be less than 10 seconds.
- Active optics open-loop stability - Once this coarse state of adjustment is achieved, the active optics must be able to hold the figure of the mirror during transitions from one guide star to another. It is too early to attempt a hard specification here, but it seems reasonable and feasible to require that the RMS image blur not degrade from 1 arcsec to more than 1.5 arcsec in the course of the combination of a five-minute delay and a 5 arcmin offset of the telescope, under open-loop active optics control. This requirement may have to be relaxed under conditions of higher wind.
Cassegrain focus - The throughput of the telescope is set by the reflectivity of the two mirrors and simple obscurations of the secondary, primary hole (about equal) and the secondary supports and by the reflectivity of the mirrors.
We require the reflectivity of the mirrors at > 90% (fresh). The geometrical throughput, on axis, is required to be:
- Primary aperture (706.8 m2; 100%)
- Secondary obscuration (7.1 m2; 1%)
- Secondary support trusses (16.4 m2; 2.3%)
- Inter-segment spaces (3 mm space+bevels; 0.6%)
So the total throughput at Cassegrain, after only two mirrors, will be required to be R2 x 0.96, where R is the reflectivity. Table 7 shows reasonable expectations for coating reflectivities, by wavelength range:
|Reflectivity (fresh coatings)|
|0.33-0.4 µm||0.4-0.7 µm||0.7-1.1 µm||>1.1 µm|
|Table 7 Mirror coatings.|
Prime focus - The prime focus is equipped with a corrector, atmospheric dispersion compensator (ADC), and central baffle for which the total throughput will be 66% with a goal of 78% in the visible.
Stray Light Control - The telescope will require baffling to restrict light from the moon or bright stars near the optical axis from falling on the focal plane.
Many of the scientific opportunities for the GSMT involve observations that will be affected by thermal radiation from the telescope, so it is important to consider the reasonable limits that can be placed on the emissivity. Although the spaces between the segments will be emissive, and the secondary support will be large, the area involved is still a small fraction of the aperture, so the use of low-emissivity coatings on the mirrors will have a significant impact on performance. For the Cassegrain focus, it appears that the throughput costs of baffling the secondary support struts and secondary obstruction outweigh, in net performance, the increased background owing to these objects in the beam. With the aggressive assumption that we will provide high reflectivity coatings, the emissivity in the near IR will be at worst 1-0.88 = 12%, with a goal of 1-0.94 = 6% if the high reflectivity values in Table 7 are met.
As mentioned above, AO involves some optical complexity. By making the secondary adaptive, a significant step has been taken toward minimizing the throughput penalty that compensating elements in the optical chain incur. Inevitably there is a tradeoff to be made between good AO performance and good AO throughput, and it is pointless to set a requirement before making it. The challenge is to aggressively minimize the negative impacts of the complexity. In the section on requirements flowdown, the results of such a particular tradeoff are presented for the multi-conjugate adaptive optics (MCAO) system. Throughputs in excess of 75% are achieved throughout the near-IR spectral range.
Eliminating an extra adaptive element also helps minimize the emissivity of the optical train feeding an AO instrument working in the IR. Beyond that the tradeoff arises again, now between improved AO performance (for example in MCAO, to achieve wider field) and worsened performance in the thermal IR. As we will see in the discussion of our example suite of instruments, there are several different ways to make the tradeoff with this telescope. At one extreme in particular, in the case of mid-IR spectroscopy, where the requirements of AO are least demanding and the thermal background most intense, we can achieve very high Strehl AO, with only two warm mirrors (and a beamsplitter, which can be cooled) to minimize thermal contamination.
A number of applications for the GSMT have been identified which require precise astrometric capabilities. Using MCAO, GSMT will achieve images with PSFs at 1.65 microns that are about 10 milliarcsec in FWHM. Current astrometric telescopes achieve precision of narrow-field differential astrometry of about 1 milliarcsec starting with PSFs of order 0.5 arcsec FWHM. Simple scaling suggests that one might hope for centroid positions precise to 20 microarcsec. Unfortunately, the limits are likely to be set not by the precision of PSF centering, but by other effects of atmospheric and instrumental origin.
The atmospheric limit is related to the differential motion of stellar images in the field caused by the difference between light paths in high turbulent layers, as first computed by Lindgren.4 The variance of the position depends on the field size. It decreases with aperture diameter and integration time as D-2t-1. For a 30-m aperture, a field of 1 arcmin radius and an integration time of 1 hour, with average seeing (r0 = 10cm) and typical turbulence and wind velocity profiles, the RMS position error is 0.035 milliarcsec. The effects of a typical finite outer scale of 25 m, which rolls off the amplitude of the most serious low spatial frequency correlations in phase, can reduce this estimate by about a factor of 10! Because of its large aperture, GSMT has enormous astrometric potential, rivaling space astrometry for narrow-field applications.
In practice it is extremely difficult to achieve atmospherically limited precision. Pravdo et al.5 at Palomar have achieved the atmospheric limit in short-term (non-AO) measurements with a precision of 0.1 milliarcsec. This was accomplished only after elimination of differential chromatic refraction, arising from uncalibrated color differences between target and reference stars. Long-term telescope stability was apparently not good enough to extend this level of precision to longer time scales. A larger telescope like GSMT, with its complex control system, will be even more challenging in this respect. Adaptive optics itself compensates part of the atmospheric errors, but introduces its own systematic effects discussed below. During exposure, the plate scale must be constant to a high degree simply to maintain image quality. Over a 2-minute field, with DL images at 1.6 microns, the scale has to be maintained fixed to 4 x 10-6 (0.5 milliarcsec) during an exposure to maintain the FWHM to within 5% of the DL width.
Figure 1 shows schematically a mapping between stars in the sky and their images on the detector, which is affected by the atmosphere (both turbulence and refraction), telescope optics, adaptive optics, and the scientific instrument. The net effect in a long-exposure astrometric frame is described by a distortion map containing both low-order and high-order components. The lowest orders correspond to scale changes, anamorphic distortion, and field rotation. These can be measured and modeled with only 3 reference stars in the field. More stars are needed to model higher orders. On the other hand, there are no stars whose positions on the sky are stable at the level of a few microarcseconds. This means that astrometric solutions can only be found in a global way, by simultaneously modeling the individual stellar motions and distortions of each frame for the whole dataset (as in space astrometry, e.g., Hipparcos). Further studies will show to what level instrumental distortions can be compensated in typical observing programs requiring precise differential astrometry. There will be an optimum balance between the complexity of astrometric models, instrument stability, and the achievable precision.
Low-order field distortions are by far the most significant terms in differential astrometry. They are produced by the atmosphere, but also by all other components in the light path, down to and including the detector. Specifically, an MCAO system generates field distortion when it applies corrections to the atmospheric aberrations of focus, and the two components of astigmatism, derived from NGSs, while compensating turbulence in a closed loop; these terms are applied to the DMs conjugate to finite altitudes. Instantaneous compensation will be very good-with full widths down to the diffraction limit-but systematic residuals will remain as field distortion that will not be averaged out in a long exposure. This is why a posteriori calibration of each field will be unavoidable.
In an MCAO system using LGSs, the low-order distortions are controlled by tip-tilt signals from three NGSs in the field. If the initial positioning of guide probes contains small errors, these errors will be nulled in closed loop by the appropriate deformations of DMs. The system will place NGSs exactly on the guide probes, and in doing so it will introduce a low-order field distortion! A special procedure has been developed to correctly position the guide probes prior to closing the loop. However, the achievable accuracy will still be orders of magnitude less than the potential accuracy of GSMT astrometry, for a number of reasons (insufficient averaging of atmospheric terms, mechanical positioning accuracy, flexure, etc.). Moreover, the NGSs themselves will move over time.
The table in Figure 1 gives a very rough idea of the relative contribution of different system components to the final distortion map and of their temporal character. Atmospheric tilt anisoplanatism that remains after long-exposure averaging is taken to be 1 unit; it defines the ultimate limit. Higher-order atmospheric terms will be much less. On the other hand, overall variations of plate scale caused by telescope and detector will be huge compared to this limit. An MCAO system will contribute both plate scale variations and astigmatic low-order terms, as explained above, at a level that clearly exceeds the atmospheric limit. All these low-order terms need to be accounted for in the process of final data modeling.
The structure of the compensated PSF and optimum techniques of extracting astrometric information from AO-compensated images are other issues that need further study. We have identified a number of questions about astrometric performance that require answers before we can realistically set targets for performance. The task of answering these questions will require integrated modeling of the interaction between the telescope and the atmosphere. We propose to pursue this task in a follow-on study in the coming year.
The GSMT will be an expensive telescope to build and to operate. Increasing its efficiency of operation is as powerful a tool for increasing its overall productivity as, for example, increasing its aperture. Yet the danger is to treat this subject as self-evident-therefore not requiring analysis-and of lower urgency.
We frequently define efficiency in terms of the fraction of the time between sunset and sunrise, when photons are being gathered by the detectors from science target sources in the sky. This must be balanced against dollar cost. In designing and costing the telescope, it is important to consider the operational model not only as it affects the user, but also as it impacts the productivity of the team of engineers and technicians who maintain the telescope.
Routine, frequently performed operations, such as acquiring calibration stars, focusing, and tuning the wavefront control systems for optimal image quality will have to be completely automated to gain the full potential of the telescope. On-telescope instrument calibration (flat-fielding, for example) is another task which must be done efficiently and rapidly so as to minimize impact on both observing time and maintenance time.
Similarly, there is potential for improving overall productivity by minimizing downtime and increasing the efficiency of maintenance.
In large measure because of its segmented nature and unprecedented dependence on active and adaptive control systems, the GSMT is more vulnerable to failure modes that arise from its complexity. But the combination of modularity, redundancy, and extensive automation of monitoring should, with sufficient up-front investment, enable the system to be robust and economical to operate and maintain. In the design process, we should set the goal of finding the point at which investment in automation reaches the point of diminishing returns in efficiency, measured in terms of total productivity per life cycle cost. At a level short of that, the investment in the telescope will be underutilized.
What are reasonable bottom-line operational goals? Recent data from the VLT (VLT Operations Report; Dec. 2000) can be used as a real-world example.
Typical data over the past few semesters are:
- Weather downtime: 12-15%
- Technical downtime: 5-7% (apparently improving as telescopes mature)
- Science efficiency: 70-90% (ratio of science integration time to available nighttime after subtraction of downtime)
These are encouraging figures and are already sufficiently close to ideal performance so that the returns from a sizeable increase in effort will not be likely to justify the cost. Using these data as a guideline to what is achievable, we propose the following goals for efficiency:
- Weather downtime: < 15%
- Technical downtime: < 7%
- Science efficiency: > 80%
- Racine, R.; Walker, G. A. H.; Nadeau, D.; Doyon, R.; Marois, C. "Speckle Noise and the Detection of Faint Companions". PASP, 111, 587 (1999).
- Veran, J.-P., Rigaut, F., Maitre, H., Rouan, D. "Estimation of the adaptive optics long-exposure point-spread function using control loop data". JOSA A, 14 (11), 3057 (1997).
- Veran, J.-P., Jolissaint, L. "Automatic reconstruction of the MCAO PSF from real-time loop data", in "Beyond Conventional Adaptive Optics", Robert Ragazzoni et al. eds., to be published by ESO, (2001).
- Lindgren, L. "Atmospheric limitations of narrow-field optical astrometry". A&A, 89, 41 (1980).
- Pravdo, S.H. & Shaklan, S.B. "Astrometric Detection of Extrasolar Planets: Results of a Feasibility Study with the Palomar 5 Meter Telescope". ApJ 465, 264 (1996).