13.6 GMACS – The GMT Wide-Field Optical Spectrograph
The wide-field optical spectrograph for GMT is called GMACS by analogy with IMACS, the
wide-field optical spectrograph for Magellan. The current concept for GMACS provides
complete, simultaneous spectral coverage over the wavelength range from ~0.36 to 1.02μ, for
hundreds of objects in an 8 x 18 arcmin field of view. The resolution with a 0.7 arcsec slit is
1400 in the blue and 2700 in the red. The optical design of the GMACS collimator, the widefield corrector described in Section 6.3.1, and the GMT secondary mirror have all been
developed together as a system. The instrument is modular and lends itself to a staged
deployment with an initial set of capabilities which can be expanded to the full set as time and
resources permit.
13.6.1 Multi-Slits or Fibers?
Two approaches are commonly used to take many spectra simultaneously over a large area of the
sky. The first is to use a multislit spectrograph and an aperture plate. The second is to use a
fiber spectrograph and a fiber positioner. The fiber spectrograph is subject to fiber losses and
generally will have somewhat lower efficiency per object than the multislit spectrograph. On the
other hand the multislit spectrograph must cope with a semi-random distribution of dispersed
images over the field of view, which typically results in a somewhat lower multiplex advantage.
High-accuracy techniques for background subtraction, for example nod-and-shuffle (Glazebrook
and Bland-Hawthorn 2001), appear to be equally applicable to either approach.
Since the emphasis in GMT is on the collecting area for seeing-limited observations (and on the
diffraction limit for AO observations), the multislit approach has been selected as the baseline in
order to avoid compromising the ultimate magnitude limit of the spectrograph. However the
collimator design for the multislit spectrograph may also be suitable for use with fibers, and an
optional fiber input appears to be an interesting upgrade path for future study.
13.6.2 Field of View
The desirable field of view for a seeing-limited spectrograph falls into the category of “more is
better.” The multiplex advantage will go up in proportion to the total slit length and there will
always be some programs which can utilize this advantage, whether because they aim to cover
the maximum possible area on the sky or because they concentrate on objects of a particular size.
The real question is what is the maximum practical field of view.
For a wide-field collimator, optical solutions based on reflecting optics tend to suffer from the
problem that the exit pupil is inconveniently located. Creating an external pupil in a Schmidtlike optic usually requires large obscurations or an off-axis arrangement, neither of which is
desirable. Transmitting optics, on the other hand, can in certain circumstances be corrected for
extremely wide fields of view while maintaining good color correction, high throughput, and an
accessible exit pupil. In the IMACS design (Bigelow and Dressler 2003), the secondary focal
ratio is optimized in order to match the focal plane curvature of the telescope to the focal plane
curvature of the transmitting collimator. The choice of a Gregorian rather than a Cassegrain
secondary is required to obtain the desired sign of the field curvature. The optical performance
13.6-1
of the IMACS collimator is very high compared to other collimators which do not take
advantage of this arrangement.
When scaling a given spectrograph design, the sky coverage and wavelength resolution depend
on the diameter of the collimated beam. Wide-field spectrographs for the current generation of
large telescopes have been built with collimated beam diameters up to 200 mm (e.g. Binospec on
the MMT, Fabricant et al. 2003). On GMT the much larger entrance aperture makes even a 300
mm beam spectrograph relatively modest in size, roughly equivalent to a 100 mm spectrograph
on an 8-m telescope, for example GMOS on Gemini (Crampton and Murowinski 2004). The
availability of most transmitting optical materials is excellent in diameters up to about 200 mm,
but gradually becomes more questionable in the range from 300 to 500 mm. A baseline
collimated beam diameter of 300 mm has been adopted for GMACS. Because the collimator and
camera optics must cover appreciable fields of view, the optical elements (except for the grating)
are usually somewhat larger than the collimated beam diameter. With modest amounts of
vignetting, the optical elements in GMACS are typically about 400 mm in diameter. The
availability of optical materials in the required sizes is being investigated as part of the GMACS
design study.
Although the diameter of a circle which fully encloses all 7 GMT primary mirror segments is
25.4 m, the optical design of the spectrograph has been carried out using a 24.4 m “reference
pupil” which is somewhat more representative of the actual illumination within the telescope
aperture. The reference pupil encloses 97% of the collecting area of GMT, but has 92% of the
area of a 25.4 m circle. For a 300 mm collimated beam, the angular magnification between the
sky and the grating is 24.4/0.3 = 81.33.
IMACS-style collimators with 300 mm diameter collimated beams can be designed for GMT
with fields of view up to about 15 arcmin in diameter. There are three regimes of increasing
complexity and cost which can be considered in order to cover increasingly large fields of view:
(1) For fields of view less than about 10 arcmin in diameter, the images in the GMT focal plane
are adequate even without a corrector (rms image diameter <0.14 arcsec, see Section 6.2.2). A
10 arcmin IMACS-style collimator would introduce very little additional aberration to the
images. The angular field of view of the collimator would be 0.167*81.33 = 13.6 deg and is
consistent with practical fields of view for spectrograph cameras, which will be discussed below.
For the simplest case without a field corrector or atmospheric dispersion compensator, it would
be necessary to align the multislits approximately along the direction of atmospheric dispersion.
The use of “microslits” (small round or square apertures) for beam switching or to sample the
highest density of objects on the sky would be possible only when working very close to the
zenith or over a limited wavelength range.
(2) For fields of view between about 10 arcmin and 15 arcmin, a two-element spherical corrector
for the telescope is required. The improvement in image quality in the telescope focal plane is
dramatic, from ~0.3 to 0.04 arcsec rms diameter. An IMACS-style collimator for GMT, with a
300 mm diameter collimated beam and a 15 arcmin diameter field of view, is shown in Figure
13.6-1. The angular field of view of the collimator is 0.25*81.33 = 20.33 deg. This collimator
has been optimized over the wavelength range from 0.35 to 1.00μ and produces polychromatic
images 0.10 arcsec in diameter (rms) everywhere in its field. However, the angular field of this
13.6-2
collimator is comparable to the practical fields of view for spectrograph cameras, and it would be
difficult to effectively utilize the available field of view while also providing significant spectral
coverage.
Figure 13.6-1. An IMACS-style collimator for GMT with a 15 arcmin field of view. The telescope focal plane is at
the left, followed by a large fused-silica field lens. The next two elements are also fused silica, followed by a
BAK2/CaF2 doublet. All surfaces are spherical. The 300 mm diameter exit pupil is shown at the right.
(3) For fields of view between 15 arcmin and 20 arcmin in diameter, a “telecentric” corrector of
the kind described in Section 6.3.1 would permit multiple collimators and spectrographs to be
deployed across the telescope focal plane in a fly’s-eye arrangement. This is the most complex
and expensive of the three alternatives, but by most measures it will also be the highest
performance. The fly’s-eye approach has been baselined for further study. The intent is to
explore just how complex and expensive this alternative will be, and to decide at a later time
whether to undertake a phased deployment of multiple spectrographs, or to fall back to a simpler
initial design.
13.6.3 Resolution and Spectral Coverage
GMT has many times the collecting area of an 8-m class telescope, and objects observed at the
spectroscopic limit will be extremely faint compared to sky. Subtraction of sky background will
depend on achieving high systematic accuracy, for example by dithering the image on the slit
between exposures or by the nod-and-shuffle technique.
There are two distinct regimes for the detection of very faint objects. In the “blue” (0.35 to
0.70μ) there are relatively few night sky lines and the background is more or less continuous. As
long as the sky background safely exceeds the detector noise, the quality of the result will not
depend strongly on resolution. In the “red” (0.70 to 1.00μ) there are many bright night sky lines.
At sufficiently high resolution, the spectral coverage between the lines dominates, and for at least
some observations the successful strategy will be to cut out the lines and either leave gaps in the
spectra where the S/N ratio is very low, or interpolate over them.
The dispersion required to “resolve out” the lines in the red will depend to some extent on the
details of a specific observation. The general impression of red spectra taken at Magellan is that
a resolution of 1800 is too low, and about twice that resolution will be required in order to open
up the area between the lines. A resolution of 3600 corresponds to a FWHM of 83 km/s. At this
resolution interpolating over the gaps in the spectra of faint galaxies, for example, begins to
make sense when the velocity dispersion is larger than 83/2.5 = 33 km/s.
13.6-3
Consider a reference slit width of 0.7 arcsec (in typical Magellan seeing, 1.0 arcsec is larger than
necessary). With an angular magnification of 81.33, a 0.7 arcsec slit corresponds to 0.0158 deg
at the grating. At the center of the red band (8500A) and a resolution of 3600, the slit width is
8500/3600 = 2.36A. To cover the complete red spectral region from 0.70 to 1.00μ, the angle
subtended by the spectrum will be 0.0158*3000/2.36 = 20.1 deg.
The very different requirements for efficient background subtraction in the red and blue parts of
the spectrum suggest that it might be worthwhile to consider making each arm of the fly’s-eye a
double spectrograph with separate red and blue cameras. There are a number of other
advantages to this arrangement. The optical design of each camera can be optimized over a
narrower wavelength range, resulting in higher optical performance (for example better image
quality, larger field of view or faster focal ratio). Anti-reflection coatings, diffraction gratings,
and detector arrays can each be optimized for the red and blue regions of the spectrum, resulting
in higher efficiency. Finally, the spectral coverage of two cameras is up to twice as large as it is
for one camera. This is true even at the very lowest resolutions where a diffraction grating
cannot cover more than a factor of two in wavelength without encountering overlap between first
and second order. These advantages are substantial and the double-spectrograph arrangement
has been baselined for further study.
13.6.4 An Imaging Spectrograph?
Transmitting gratings are the preferred dispersing elements for GMACS. These may be either
surface-relief or volume-phase holographic (VPH) gratings, either with or without auxiliary
prisms to steer the diffracted beam in a preferred direction. Although low-order reflecting
gratings can achieve somewhat higher angular dispersion than conventional transmitting
gratings, they have little advantage in dispersion compared to VPH gratings. Transmitting
grisms typically require less space between collimator, grating and camera, and they exhibit
lower anamorphic distortion, both of which simplify the challenge of designing the spectrograph
optics. Finally the blaze efficiency for VPH gratings is substantially higher than for reflecting
gratings.
For a true imaging spectrograph which can take an undispersed image of the sky, the
spectrograph camera must be aligned with the optical axis of the collimator. When a
transmitting grating is introduced into the beam, the diffracted light must either be redirected into
the camera by using a prism, or the camera must mechanically articulate away from the optical
axis of the collimator. The limited angular adjustment available with prisms limits the
achievable dispersion of transmitting gratings, while multiple articulating cameras of unusually
large size are a daunting prospect in terms of mechanical complexity and stability.
An alternative is to abandon the ability to make an image of the sky through the spectrograph.
The cameras can be permanently fixed at an intermediate angle where higher dispersions can be
achieved by introducing prisms which deflect the light in one direction, and lower dispersions
can be achieved by introducing prisms which deflect the light in the opposite direction. One
consequence of this choice is that alignment of objects onto the multislits must be done by using
separate small cameras at the aperture mask, similar to the way that many fiber systems are
aligned. A possible way to preserve the mask alignment capability through the blue camera is
discussed in Section 13.6.8.
13.6-4
Another consequence of this choice is that the imaging requirements for GMT must be satisfied
with a separate instrument devoted exclusively to imaging. There are advantages to this
arrangement for a variety of reasons. Spectrograph cameras must be optimized to cover a wider
field of view than an equivalent reimager, because of the need to accommodate the angular
dispersion of the grating. An imaging instrument can be optimized to correct the aberrations in
the telescope focal plane, and atmospheric dispersion (for seeing-limited observations through
typical broad-band filters at reasonable air mass) is usually not a problem, so a reimager can
dispense with the corrector/ADC. Finally, whereas a spectrograph must be optimized to provide
an accurately collimated beam at the grating, a reimager requires only an approximately
collimated beam, for example for use with a Fabry-Perot. The result is that it should be possible
to design a reimager which is equivalent to at least one arm of a fly’s-eye spectrograph, but with
a considerably simpler and more efficient optical system. Such a system might be similar to the
Four-Star instrument currently being built for Magellan, but re-optimized for use at optical rather
than near-IR wavelengths.
One goal in designing the area around the instrument platform on GMT is to provide for
relatively straightforward exchange of instruments at the Gregorian focus. If this goal can be
met then the prospect of implementing a separate imaging instrument becomes at least somewhat
more palatable. An optical design for such a reimager is a necessary ingredient in order to
evaluate the advantages and disadvantages of this arrangement. Pending further study, the nonimaging configuration with separate red and blue cameras at fixed intermediate angles has been
selected as the baseline concept for the GMT spectrograph.
13.6.5 GMACS Fly’s-Eye Collimators
The “telecentric” field corrector/ADC described in Section 6.3.1 provides a 20 arcmin diameter
field of view. Four fly’s-eye collimators, with fields of 8 x 8 arcmin each, would make efficient
use of the corrected field, providing a 16 x 16 arcmin overall field of view with small lost areas
at the corners where they are cut off by the 20 arcmin circle. The total available slit length
would be 32 arcmin.
For each collimator, the angular field of view at the grating would be 8*81.33/60 = 10.8 x 10.8
deg. The diagonal of the square is 15.3 deg. In Section 13.6.3, the angular dispersion for
complete spectral coverage of the red region at adequate resolution is calculated to be 20.1 deg.
In order to provide complete spectral coverage for any object in the 8 x 8 arcmin field, the
angular field of view of the red camera would have to be 10.8+20.1 = 30.9 x 10.8 deg. The
diameter of the camera field of view would have to be 32.7 deg. This is an extremely wide field
for a spectrograph camera. The camera in the DEIMOS spectrograph at Keck (Epps 1998) has a
23 deg field of view. The IMACS short camera (Epps and Sutin 2003) has a field of view of
about 27 deg. Both of these cameras are designed for 150 mm spectrographs (although the
DEIMOS camera accepts a substantial anamorphic beam expansion from its reflecting gratings),
and it is not clear whether it would be practical to scale either camera up by the required factor of
two in size.
There are three ways to conserve the available field of view of the camera. The first is to adopt
an even larger collimated beam diameter, but 300 mm has already been selected to be the largest
practical size, given the constraints imposed by the feasible sizes of optical glass elements. The
13.6-5
second is to use a narrower slit. There will be some loss of efficiency but less than the factor of
two that might result from taking two separate exposures with higher dispersion gratings. The
third way is to restrict the field of view in the dispersion direction.
If VPH gratings are used, it may be necessary to restrict the field of view in the dispersion
direction anyway. For VPH gratings there is a large variation of the blaze function (and in
particular the blaze wavelength) with incident angle. As will be discussed below, a field of view
on the sky of about 4 arcmin in the dispersion direction appears to be the upper limit for VPH
gratings. The greater efficiency of VPH gratings could at least partially compensate for the loss
of sky coverage. VPH gratings also provide access to higher dispersions than surface-relief
transmission gratings, and can be supplied in one piece rather than as the mosaic that would most
likely be required for a 300 mm surface-relief grating.
For these reasons the GMACS fly’s-eye field of view has been divided into four 4 x 9 arcmin
quadrants. There are small gaps between quadrants, so the overall field will be about 8.5 x 18.5
arcmin in size. The diagonal of 20.4 arcmin is only very slightly larger than the 20 arcmin field
of the corrector/ADC. Compared to the 8 x 8 arcmin quadrants, the total available slit length has
been increased from 32 to 36 arcmin, but the field area has been decreased from 250 to 144
arcmin2. The decision to adopt a rectangular rather than a square field of view should definitely
be revisited after further study of the properties and trade-offs of VPH and surface-relief
gratings.
The design of the collimator for each quadrant starts out with the baseline GMT, with field
corrector and ADC, and is optimized with the collimator positioned on the optical axis. The
optimization is easier to perform on a fully-axial system which also requires taking out the prism
tilts in the ADC, but the change in the images compared to the tilted ADC in the “cancelling”
(zenith) position is negligible. The optical layout is shown in Figure 13.6-2.
Figure 13.6-2. Axial layout for the GMACS fly’s eye collimator. The last element of the GMT field corrector is
shown at the left, and serves as the field lens for the collimator. The corrector is followed by the telescope focal
plane, which is 9.85 arcmin in diameter (the diagonal of a 4 x 9 arcmin field). Following the focal plane, the first
two elements of the collimator are fused silica, followed by a BaK2/CaF2 doublet. The exit pupil is shown at the
right.
13.6-6
The spot diagram for the GMACS collimator in the on-axis position is shown in Figure 13.6-3.
In order to form an image, a paraxial (perfect) camera element has been appended at the exit
pupil of the collimator, with a scale of 982 μ/arcsec. The rms image diameter is 0.07 arcsec.
The collimator has been optimized for all wavelengths between 0.35 and 1.0μ and lateral color is
negligible.
Figure 13.6-3. Spot diagram for GMACS collimator, on axis. The box size is 0.25 arcsec.
To work as a fly’s eye, the collimator must be rotated off axis around the center of curvature of
the focal plane. The rotation is the only change in the optical prescription – no additional
optimization is performed. The rotated configuration is shown in Figure 13.6-4. In order to
prevent physical interference between multiple collimators, it is necessary to introduce a
diagonal mirror for each quadrant. The diagonal mirrors are located close to the focal plane and
are discussed in a subsequent section.
13.6-7
Figure 13.6-4. Fly’s-eye collimator rotated into an off-axis position.
The spot diagram for the off-axis collimator is shown in Figure 13.6-5. Since rotating the
collimator off-axis breaks the axial symmetry of the system, spots are shown for 5 equallyspaced field radii, starting from a point near the center of the corrected field, and ending with a
point near the edge. The typical rms spot diameter is still about 0.07 arcsec, except near the very
edge of the field where it is less than 0.10 arcsec.
Figure 13.6-5. Spot diagram for GMACS collimator, rotated to an off-axis position. The box size is 0.25 arcsec.
13.6-8
The final element in the GMACS collimator is made of calcium fluoride (CaF2) and has been
limited in diameter to 370 mm. This limitation introduces a moderate amount of vignetting
which is shown in Figure 13.6-6. The maximum value for the vignetting is 15%, which occurs at
the extreme corner of the field (where the increment of area dA/dr falls to zero). The average
light loss for an ensemble of objects distributed randomly across the field would be about 4%.
The alternative way to minimize the diameter of the spectrograph optics is to reduce the size of
the collimated beam. In order to maintain the same resolution, the slit width for all objects
would have to be reduced in direct proportion, leading to much more serious losses than is the
case for a modest amount of vignetting.
Figure 13.6-6. Vignetting diagram for GMACS fly’s-eye collimator.
The diameter of 370 mm is the upper limit for high-quality CaF2 material supplied by Schott,
which supplies the bulk of the CaF2 used worldwide to fabricate optics for semiconductor UV
lithography. According to Fabricant, high-quality CaF2 is also available in diameters up to 400
mm from Canon, and CaF2 is available up to 530 mm in diameter from Bicron/Saint Gobain.
The Bicron material is typically multicrystalline, but multicrystalline material has been used
successfully to produce the optics for many astronomical instruments.
The last surface of the CaF2 collimator lens is also aspheric. Unlike glass, calcium fluoride can
be single-point diamond turned to an aspheric shape at only a modest cost increment compared
to a polished spherical surface. The surface quality produced by the newest and best machines is
good enough that no additional post-polishing is required in order to reduce surface ripple and
scattered light.
Figure 13.6-7 is a profilometer trace of a 150 mm diameter Schmidt corrector plate recently
made by Janos Optical for the Magellan echellette spectrograph. The trace shows the deviation
from the desired shape, which is an asphere superposed on an overall meniscus with a sag of 4
13.6-9
mm. On centimeter scales, the maximum slope error of the surface is 5 arcsec. In the Magellan
echellette, the angular magnification between the telescope aperture and the Schmidt camera is
65 (similar to the factor of 81.33 for GMACS). There is an additional factor of 2.3 which is the
ratio between the surface slope error and the refracted angle error in a material with an index of
refraction of 1.43. The result is that the geometric ray error which results from the slope error
corresponds to an angle on the sky of 5/(65*2.3) = 0.033 arcsec. Since this is the maximum
rather than the rms slope error it is somewhat more conservative than an rms spot radius. An
equivalent rms spot diameter might be about 0.05 arcsec, which is acceptable, although a factor
of 2 improvement would be even better.
Figure 13.6-7. Surface profile of CaF2 asphere.
There are a few small-scale bumps in the surface profile. For the most part they repeat at the
symmetrical location across the center of the trace, so they correspond to complete circular zones
with a very small scale. A detail of the surface profile near the worst of these bumps is shown in
Figure 13.6-8. The amplitude of the bump is 30 nm, and the corresponding wavefront error is 10
nm. The more typical rms surface ripple is about 2 nm and the rms wavefront ripple about 1 nm.
The overall scattering from this surface, even at UV wavelengths, should be very low. Although
the original contract called for post-polishing the surface, the decision based on the profilometry
data was to skip the post-polishing step. The same machine has the capacity to machine surfaces
up to 425 mm in diameter.
13.6-10
Figure 13.6-8. Detail of surface profile. The lower points are the individual readings with a spacing of 5μ, and
probably contain 1 or 2 nm (rms) of measurement noise. The upper trace is the median value in a 100μ interval,
with an offset added for clarity.
One final note is that the GMACS collimator design calls for a doublet made of glass and
calcium fluoride. The difference in CTE between the two materials, and the fragility of CaF2,
implies that the two elements cannot be glued together in the conventional way. There are 3
options for making this doublet. (1) The surfaces can be glued together using a thick layer of
compliant cement, like Sylgard 184 RTV. This method has been used extensively on the MIKE
spectrograph at Magellan (Bernstein et al. 2003), where the lenses are typically 220 mm in
diameter and the glue layer is 250μ thick. For the GMACS collimator, the glue layer might have
to be 400 - 500μ thick. The MIKE lenses were glued at a reduced temperature (15 ºC) in order to
minimize the differential expansion over the working temperature range at the telescope. (2) The
lenses could be oiled together, as has been done extensively for IMACS and other spectrographs.
Generally the use of oil requires some precaution to minimize axial motions of the elements
caused by changes in the hydrostatic pressure of the oil as a function of position, and a
compensating reservoir to take up changes in the volume of the oil as a function of temperature.
(3) The two elements in the doublet could be separated slightly and the optical design reoptimized to take advantage of the extra degrees of freedom. Since it is likely that a very high
performance AR coating will be used (e.g. sol-gel over MgF2, at least on the interior surfaces of
the collimator) the loss of efficiency would be only 1-2%, depending on wavelength.
13.6.6 GMACS Red and Blue Cameras
At the present time it seems unlikely that ground-layer AO will be a factor at optical wavelengths
over the 20 arcmin field of view of the GMACS fly’s eye. However there will be times of good
seeing which the spectrograph optics should not degrade significantly. A reasonable goal would
13.6-11
be to design the spectrograph optics for 0.40 arcsec FWHM seeing. If the telescope, corrector,
collimator and camera taken together, including their design aberrations, fabrication errors and
alignment tolerances, produce final rms image diameters of 0.25 arcsec, the final image quality
in 0.40 arcsec seeing will be degraded to 0.47 arcsec. It will be a challenge to meet a 0.25 arcsec
error budget, and a tighter specification is probably unrealistic.
The rms image diameter delivered by the Magellan telescopes, as measured by their active optics
systems, is 0.12 arcsec. The rms image diameter of the GMT telecentric corrector design with
ADC is ~0.05 arcsec. Following the discussion in Section 6.3.5, about 0.05 arcsec (actually .025
arcsec in several places) must be added for alignment errors in the corrector. The rms image
diameter for the collimator design is ~0.08 arcsec. It is too soon to make a rigorous estimate of
the fabrication and alignment errors for the collimator and camera, but an optimistic number
might be 0.10 arcsec. Taking all of these errors in quadrature leaves a maximum value for the
contribution of the camera design of 0.16 arcsec.
Although a more rigorous error budget will be developed after further work on both the telescope
and GMACS, a reasonable goal for the rms image diameter of the camera design is between 0.10
and 0.15 arcsec. Taken in quadrature, this is between 16% and 36% of the total available error
budget. There is good reason to allocate such a large fraction of the error budget to the camera:
desirable properties like field size, throughput and spectral coverage are likely to be enhanced
significantly at the expense of optical aberration.
The GMACS cameras should be fast enough that the entire 9 arcmin slit can be imaged onto the
length of 2 CCD’s in a 2 x n mosaic of 3-side buttable CCD’s. Depending on the exact length of
the CCD’s, the interesting range of focal ratios is between 2 and 3. Note that imaging the slit
along the mosaic in this direction is required for nod-and-shuffle operation.
While exhibiting a suitable optical design is an unequivocal way to define what is possible in an
optical system, it is much more difficult to define what is not possible. The current designs for
the GMACS red and blue cameras are the result of several months of design work, but it is likely
that better designs can be developed with additional effort. The designs evolved from a
configuration adapted many years ago from one of the cameras in the UVES spectrograph on
VLT (Dekker et al. 2000), but at this point they bear little resemblance to any original and
describing how they evolved would be difficult and not particularly enlightening.
The layout for the GMACS blue camera is shown in Figure 13.6-9. The focal length is 789 mm
(f/2.63); the scale at the GMACS focal plane is 311 arcsec/mm. All of the glasses have been
selected for high UV transmission, and the design has been optimized over the wavelength range
3700 - 6700A. There are two calcium fluoride elements that have been placed in positions which
minimize their diameters (334 and 350 mm) and that incorporate aspheric surfaces following
much the same reasoning that has been described for the collimator. The remainder of the
positive power is provided by glass type FPL51Y, including a fairly strong first element. By
concentrating the positive power near the front of the lens, the overall diameter of the subsequent
optics has been minimized (the largest element is the FSL5Y, which is 370 mm in diameter).
The angular field of view is 23.0 deg; the diameter of the focal plane is 321 mm. The field
flattener is very strong and the variation in angle of the chief ray at the focal plane is significant.
As a result there may be a small change in image scale with focus. In any event, the question of
13.6-12
how the camera focus should be adjusted (as a function of temperature, for example) requires
further study.
Figure 13.6-9. GMACS blue camera layout. The entrance pupil is at the left, with a 130 mm space to the first
element of the lens. The first doublet is FPL51Y/BSM51Y, followed by a CaF2 singlet with an asphere on the
trailing surface. The quadruple element is BAL15Y/FPL51Y/BSM51Y/FPL51Y, followed by an FSL5Y/FPL51Y
doublet and a second CaF2 singlet, again with an asphere on the trailing surface. The field flattener/dewar window
is fused silica.
The optimization of the camera allows for a modest amount of lateral color, which would be
acceptable even if the camera were to be used for imaging. The spot diagram is shown in Figure
13.6-10, and the rms image diameter as a function of field position and wavelength is given in
Table 13.6-1. The rms image diameter, averaged over all fields and all wavelengths is 0.14
arcsec.
Figure 13.6-10. Spot diagram for GMACS blue camera. The box size is 1.0 arcsec.
13.6-13
Table 13.6-1. RMS image diameter (arcsec) for GMACS blue camera.
Field/
Wavelength
0.00 deg
2.88 deg
5.76 deg
8.64 deg
11.52 deg
3700A
0.06
0.09
0.13
0.21
0.31
4200A
0.17
0.15
0.13
0.12
0.17
4800A
0.17
0.15
0.14
0.09
0.13
5500A
0.14
0.13
0.12
0.04
0.11
6700A
0.18
0.17
0.15
0.08
0.12
Once again the design incorporates a modest amount of vignetting in order to minimize the
diameter and thickness of the optics. The vignetting diagram is shown in Figure 13.6-11. The
total weight of glass which would be required to build an unvignetted version of the camera is
~1.6 times the weight of glass in the vignetted version, while the average light loss is only about
2%.
Figure 13.6-11. Vignetting diagram for GMACS blue camera.
The optical performance of the blue camera has been enhanced by keeping the CaF2 elements as
singlets, at the same time avoiding the problem of bonding glasses with very different CTE’s.
There is actually very little improvement in the image quality of the blue camera when the red
wavelengths are eliminated from the merit function. However the improvement in image quality
when the same design is optimized only for red wavelengths is dramatic. Starting from the blue
design, the red camera has been re-optimized for a significantly larger angular field (27 deg),
13.6-14
faster focal ratio (f/2.25), and fewer elements. The layout for the red camera is shown in Figure
13.6-12.
Figure 13.6-12. GMACS red camera layout. The entrance pupil is at the left, with a 150 mm space to the first
element of the lens. The leading quadruple element is FPL51/FSL5/BSM51Y/CaF2. The trailing surface of the
CaF2 element is aspheric. The center triplet is FPL51/PBL1/FPL51, followed by a CaF2 singlet, also with an
asphere on the trailing surface. The field flattener/dewar window is fused silica.
The focal length of the camera is 675 mm; the scale at the GMACS focal plane is 266 μ/arcsec.
Because the index data for one of the glasses is not specified beyond 9000A, the design is
currently optimized over the range 0.65 – 0.90μ, but it seems unlikely that there will be a
problem in extending the wavelength range to 1.0μ. Once again the two calcium fluoride
elements have been placed in positions which minimize their diameters (344 and 354 mm). The
first CaF2 element is in contact with the preceding lens, but the CTE problem can be avoided if
necessary by introducing a small space between the two. Because of the wider angular field of
view, the center triplet is somewhat larger (404 mm) than the largest lenses in the blue camera.
The spot diagram is shown in Figure 13.6-13, and the rms image diameter as a function of field
position and wavelength is given in Table 13.6-2. The rms image diameter, averaged over all
fields and all wavelengths is 0.13 arcsec.
13.6-15
Figure 13.6-13. Spot diagram for GMACS red camera. The box size is 1.0 arcsec.
Table 13.6-2. RMS image diameter (arcsec) for GMACS red camera.
Field/
Wavelength
0.00 deg
3.38 deg
6.75 deg
10.14 deg
13.50 deg
6500A
0.16
0.15
0.07
0.08
0.20
7000A
0.17
0.15
0.07
0.06
0.18
8000A
0.19
0.17
0.09
0.05
0.16
9000A
0.20
0.19
0.11
0.05
0.15
The red camera design also incorporates a modest amount of vignetting in order to minimize the
diameter and thickness of the optics. The vignetting diagram is shown in Figure 13.6-14.
13.6-16
Figure 13.6-14. Vignetting diagram for the GMACS red camera.
13.6.7 GMACS Red and Blue Gratings
Although the “red” regime of bright night sky lines begins at about 7000A, the difficulty of
maintaining high grating blaze efficiency over the blue wavelength band (which covers nearly a
factor of 2 in wavelength) makes it advantageous to shift the crossover wavelength to ~6500A.
The current GMACS baseline red grating has 790 lines per mm and works from 6500A to
10200A. The full angular dispersion is 17.9 deg. Adding the 4 x 9 arcmin field of view on the
sky results in a total field of view at the camera of (17.9 + 5.4) = 23.3 x 12.2 deg, with a diagonal
of 26.3 deg, slightly less than the 27.0 deg field of view of the red camera. The angular
dispersion could be increased from 17.9 to 18.7 deg, but is still less than the angle
20.1*3700/3000 = 24.8 deg which results from scaling the value derived in Section 13.6.3 for the
increased red bandpass in the baseline design. The implication is that a slit width of ~0.5 arcsec
will be required to achieve a spectral resolution of 3600 in the red while maintaining complete
spectral coverage. The spectral resolution with a 0.7 arcsec slit would be 2700 at 8500A.
The current GMACS baseline blue grating has 720 lines per mm and works from 3600A to
6700A. The full angular dispersion is 13.0 deg. Adding the 4 x 9 arcmin field of view on the
sky results in a total field of view at the camera of (13.0 + 5.4) = 18.4 x 12.2 deg, with a diagonal
of 22.0 deg, somewhat less than the 23.0 deg field of view of the blue camera. The angular
dispersion could be increased from 13.0 to 14.1 deg. The resulting spectral resolution with a 0.7
arcsec slit would be 1400 at 5000A.
A preliminary analysis of the performance of the baseline red and blue gratings, if implemented
using VPH technology, has been performed using the program GSOLVER. The resulting blaze
13.6-17
efficiencies as a function of field angle on the sky are shown in Figures 13.6-15 and 13.6-16.
The adopted grating parameters are listed in Table 13.6-3.
Figure 13.6-15. Efficiency curve for GMACS red VPH grating, for field angles of -2, -1, 0, +1 and +2 arcmin
relative to the field center.
The red efficiency curve at a field angle of 0 arcmin was calculated with a wavelength spacing of
10A. All of the other curves were calculated with a spacing of 100A. It is not clear what causes
the sharp features at 10A spacing. It seems more plausible that they are due to a bug in the
program than that they are physical. The parameters of the blue VPH grating result in extremely
poor performance at a field angle of +2 arcmin, so the efficiency curves have been calculated
over the range -3 to +1 arcmin instead. This is equivalent to rotating the grating and camera by
1.4 deg in order to optimize the blaze function.
Even with the field of view restricted to 4 arcmin, the losses at the extremes of wavelength and
field angle are appreciable for both the red and blue gratings. Nevertheless, there is only a small
range of wavelength and field angle for which the efficiency is less than 50% of the peak value,
and significant variations of efficiency with wavelength are typical in grating spectrographs.
Further study of the VPH grating parameters, as well as a comparison to the performance of
surface-relief gratings, is clearly required.
13.6-18
Figure 13.6-16. Efficiency curve for GMACS blue VPH grating, for field angles of -3, -2, -1, 0 and +1 arcmin
relative to the field center.
Table 13.6-3. VPH grating parameters.
Red
Blue
0.65 – 1.02
0.36 – 0.67
790
720
19.258
10.580
Thickness (μ)
5
5
Average index
1.3000
1.3000
Index modulation
0.0835
0.0510
Wavelength (μ)
Lines/mm
Bragg angle
In addition to the red and blue baseline gratings which give complete spectral coverage, the
GMACS concept calls for four additional gratings (two red and two blue) which would provide
twice the resolution but would require two exposures in order to cover the entire spectrum.
These gratings would require ~1500 lines per mm, but the required VPH grating parameters (e.g.
index modulation and gel thickness) and the performance of these gratings have not yet been
investigated.
13.6-19
13.6.8 GMACS Optical Layout
The dichroic mirror which separates the red and blue channels must be placed in the parallel
beam, between the collimator and the gratings. Attempts to put the dichroic mirror earlier in the
optical path of the collimator failed because of the aberrations produced in the red beam by
transmission through the oblique thickness of the dichroic substrate. The dichroic is tilted by 38
deg relative to the optical axis of the collimator, which should lead to somewhat improved
performance compared to an angle of 45 deg. The complete layout of a single arm of the fly’s
eye spectrograph is shown in Figure 13.6-17. For clarity, the single-arm layout is shown before
the collimator is rotated away from the optical axis of the telescope.
Figure 13.6-17. Layout of a single arm of GMACS. The last element of the “telecentric” corrector is shown at the
upper right and serves as the field lens. The dichroic mirror transmits red light to the red grating and camera, and
reflects blue light to the blue grating and camera.
The fixed angles of the cameras have been adjusted so that no prisms are required for the two
baseline gratings. There is a generous amount of space around the gratings for adding prisms to
redirect the diffracted light from higher-dispersion gratings into the cameras. The gratings
shown are 360 mm square, and are placed fairly close to the exit pupil of the collimator. The
angle of the blue camera (21 deg) is considerably smaller than the angle of the red camera (38.4
deg). It is an interesting question whether a prism could be made from crown and flint glasses
which would allow the blue camera to make an undispersed image of the slit mask over a
significant wavelength interval. This would allow the blue camera to be used to check the
alignment of objects on the slits, for which the poor UV transmission of a strong flint component
would not be an issue. The angle of the red camera is probably too large for such a prism to be
feasible.
A 3-D view of all 4 arms is shown in Figure 13.6-18. The 4 diagonal mirrors have been arranged
into an assembly called the “tent mirror” which directs light from the four quadrants of the field
into two main “optics modules” on opposite sides of the instrument platform. In the center of the
platform is the focal plane assembly which includes the field lens, aperture plate and tent mirror
13.6-20
together with guiders for maintaining the alignment of objects on the aperture plate, and image
analyzers for maintaining the active-optics alignment of the telescope.
Figure 13.6-18. 3-D view of GMACS. The “tent mirror” is shown in green.
A detail of the tent mirror is shown in Figure 13.6-19. The four facets of the tent mirror are
drawn as rectangles which are aligned parallel to the “ridgeline” of the tent. In fact the facets of
the tent mirror need to be trapezoidal in shape in order to prevent the surfaces from overlapping
near the middle of the field.
Figure 13.6-19. Detail of rays from the corrected GMT focal plane to the GMACS tent mirror.
13.6-21
The field points for the four quadrants are drawn with gaps of 30 arcsec along the ridgeline, and
gaps of 30, 45 and 60 arcsec, moving down (and further out of focus) along the sloped sides of
the tent near the center of the field. The 30 arcsec and 30 - 60 arcsec gaps are probably too
conservative, and might be reduced to 20 arcsec and 20 - 50 arcsec. Alternatively, the spacing
between the focal plane and the tent mirror could be increased somewhat in order to improve
mechanical access in this area.
13.6.9 GMACS Optical Performance
Spot diagrams for the complete GMACS optical design are shown in Figure 13.6-20 for the red
camera and Figure 13.6-21 for the blue camera. The diagrams are accompanied by Tables 13.6-4
and 13.6-5, which list the rms image diameters. The glass catalog entry which limited the
optimization of the red camera to 9000A has been modified to extend the range to 10200A,
although the accuracy of the index formula in the extended wavelength range has not yet been
verified.
Figure 13.6-20. Spot diagram for the complete GMACS optical design, with the red camera and 790 line/mm
grating. The box size is 0.5 arcsec.
Table 13.6-4. RMS image diameter (arcsec) for complete GMACS red arm.
Wavelength/Field
1
2
3
4
5
6
7
8
9
6500A
0.07
0.09
0.10
0.08
0.08
0.10
0.14
0.12
0.17
7700A
0.12
0.21
0.13
0.11
0.19
0.12
0.10
0.14
0.12
8900A
0.08
0.16
0.12
0.13
0.24
0.13
0.15
0.24
0.14
10200A
0.16
0.10
0.17
0.12
0.08
0.07
0.09
0.16
0.07
13.6-22
Figure 13.6-21. Spot diagram for the complete GMACS optical design with the blue camera and 720 line/mm
grating. The box size is 0.5 arcsec.
Table 13.6-5. RMS image diameter (arcsec) for complete GMACS blue arm.
Wavelength/Field
1
2
3
4
5
6
7
8
9
3600A
0.21
0.20
0.18
0.25
0.26
0.24
0.32
0.32
0.29
4500A
0.31
0.31
0.29
0.29
0.29
0.31
0.25
0.29
0.30
5600A
0.18
0.21
0.14
0.21
0.21
0.17
0.21
0.21
0.15
6700A
0.13
0.06
0.11
0.10
0.14
0.06
0.12
0.20
0.08
The rms image diameter for the GMACS red arm, averaged over all fields and all wavelengths, is
0.13 arcsec. This is fully consistent with the image quality goals discussed in Section 13.6.6.
The rms image diameter for the GMACS blue arm, averaged over all fields and all wavelengths,
is 0.21 arcsec. The blue image quality evidently suffers from some direct reinforcement of
aberrations in the collimator and camera (which do not necessarily have to add in quadrature),
particularly at 4500A. Possibly the blue images can be improved with further optimization. If
not the current design is at least marginally acceptable – at least the images are better in the red,
where the seeing is generally better as well.
13.6-23
13.6.10 Detectors
The 24.1 x 12.2 deg focal plane of the red camera is 289 x 147 mm in size. The 19.5 x 12.2 deg
focal plane of the blue camera is 272 x 170 mm in size. A substantial mosaic of 3-side buttable
CCD’s will be required in either case.
The sensitive area of the largest CCD in the E2V catalog is 62.21 x 27.65 mm. A 2 x 10 mosaic
of these CCD’s would approximately cover the focal plane of either camera, although the
available slit length would be only 7.6 armin for the red arm, and only 6.6 arcmin for the blue
arm. Since the number of CCD’s required by GMACS will be substantial, it seems reasonable to
ask whether a CCD with a custom-made geometry would be more suitable.
Current CCD fabrication is typically done by foundries which process 6-inch silicon wafers.
Although the wafer size is likely to grow over the next decade, the 6-inch size is proven and
probably adequate for producing the CCD’s required by GMACS. The CCD pixels must fit
within an area on the wafer which is 135 mm in diameter. This dimension already allows for a
small extra margin for the output registers, contacts, clock distribution and cutting margin
between arrays.
Table 13.6-6 lists the dimensions of the active areas of some CCD’s which would just fit within
the available area on a 6-inch wafer:
Table 13.6-6. Possible custom CCD geometries for GMACS.
# per wafer
Aspect ratio
Dimension (mm)
Red Mosaic
Blue Mosaic
2
1.62:1
85.0 x 52.4
2x6
2x5
3
2.43:1
85.0 x 35.0
2x8
2x8
2
1.30:1
73.5 x 56.6
2x5
—
3
1.95:1
73.5 x 37.7
2x8
—
From the table it is evident that there is relatively little advantage in making different CCD
geometries for the red and blue cameras. The 85 x 35 mm CCD listed on the second line is a
conservative choice. The yield should be acceptable, since the area of this CCD is only 79% of
the area of the 4k x 4k CCD’s which are becoming available from suppliers such as Fairchild and
Semiconductor Technology Associates. The 2 x 8 mosaic is intermediate in complexity between
the 2 x 4 mosaics such as the one on IMACS, and very large mosaics such as Megacam on the
MMT (Fabricant et al. 2004).
13.6.11 Flexure Compensation and Baffles
The optical components of GMACS shown in Figure 13.6-18 span the full diameter of the GMT
instrument platform. Given the large moment loads from instruments mounted both above and
below the platform, together with the change in the applied gravity vector caused by the motion
13.6-24
of the telescope in elevation and the rotation of the instrument platform, it seems unlikely that
GMACS can be made adequately stable without some form of active flexure compensation. In
order to derive the appropriate corrections in real time, it will be necessary to include a small
auxiliary CCD at the edge of each detector mosaic, and a faint comparison light source at the
edge of the aperture plate.
There are two criteria for defining adequate stability. The first arises from the requirement that
the images of the multislits should be stable enough to avoid smearing and loss of resolution
during an exposure. The second has to do with fringing on the CCD’s and the stability of flatfield calibrations.
The smearing requirement should be satisfied if the images move significantly less than the
expected image diameter. Limiting the motion to 0.1 arcsec on the sky might be appropriate.
This corresponds to a 100 μ displacement of the aperture mask, or of the tent mirrors, relative to
the optics modules. Since the focal length of the collimators is about 2.5 m, it also corresponds
to an angular change in the orientation of the optics modules of 100μ/2.5 m, or 8 arcsec.
The fringing effect is caused by interference between the incident wavefront and the wavefront
reflected by the back surface of the CCD. The index of refraction n of silicon is about 3.7. If the
thickness of the CCD is t and the wavelength is λ, then the phase difference will be 2π*n*2t/λ ≈
50t/λ . For a typical CCD thickness of 20μ, and wavelength of 0.8μ, the phase difference is
about 1200. The change in the fringing effect will be small if the change in the phase is small. If
the change in phase is 0.1, for example, then the corresponding change in wavelength is 1/12000.
For the red arm of GMACS the resolution is about 1800/arcsec, so a motion of 0.15 arcsec will
result in a phase difference of 0.1.
Figure 13.6-22 shows a small section of a flat-field image taken with the red side of MIKE, a
cross-dispersed echelle spectrograph at Magellan. In the figure, the echelle dispersion runs from
left to right, and the cross-dispersion runs from top to bottom. The wavelength varies from
~7300A at the bottom of the figure to ~9500A at the top. The fringing effect is presumably
caused by very small fractional changes in the thickness of the CCD and is weak for λ < ~8500A
because the silicon becomes opaque. The peak-to-valley amplitude of the worst fringing is about
15%.
Figure 13.6-22. Portion of a flat field exposure taken with the MIKE spectrograph.
13.6-25
For extremely precise background subtraction, or if the red CCD’s are made from specially
processed thick (~50μ) material, the stability requirement might become much more stringent,
~0.01 arcsec (or less) for 0.1% precision. In this case the use of beam-switching techniques such
as nod-and-shuffle, which do not depend on very accurate flat-field calibration, is likely to be a
more successful strategy.
Field distortions in the optics of the collimator and the camera limit the potential accuracy of an
active flexure compensation system because the image displacement will not be a constant value
across the full field of view. Field distortion usually results from having strong optical elements
close to the focal plane. The configuration of the fly’s eye collimator is favorable in this regard
since there is a large separation between the aperture plate and first element of the collimator.
The image scale at the edge of the collimator field is 0.88% larger than the image scale at the
center of the field. Tests using the optical design model demonstrate that by tilting the tent
mirror it is possible to compensate for flexure between the aperture plate and the collimator
optics module to an accuracy of 1%.
The distortion of the red and blue cameras is considerably larger. The image scale at the edge of
the red camera field is 5% larger than the image scale at the center. For the blue camera, the
image scale as a function of radius peaks at a value which is 2.4% higher than the value at the
center of the field. The most discrepant scale occurs at about 75% of the maximum field radius.
The conclusion is that there are no distortions anywhere in the optical system which would
prevent the flexure compensation system from being accurate to about 1 part in 20.
There are several obvious places to introduce flexure compensating elements. In IMACS, the
CCD platen is actuated by piezoelectric devices inside the dewar. This is not particularly
convenient because the actuators and stage must operate in a vacuum, where space is limited and
the assembly is not very accessible. The alternative is to actuate the facets of the tent mirror and
the dichroics. Since tilting the tent mirror will change the image position in both the red and
blue cameras, and tilting the dichroic will change the image position in the blue camera only, this
system is capable of controlling the image position in the red and blue focal planes
independently.
One final note concerns the optical baffles required to observe with GMACS. The intention is to
provide the main aperture stop within the instrument, at the position of the grating where the
quality of the pupil image is fairly good, although not entirely free of aberration. The difficulty
is that the telescope pupil, which is defined at the secondary, has 7 circular subapertures with sky
background showing through between them. The pattern of subapertures rotates with respect to
the instrument because of the rotation of the instrument platform.
It would be very inconvenient to provide a rotating aperture stop with 7 subapertures at each
grating. A simpler solution is to add a flat circular baffle immediately behind the secondary
mirror. The circular baffle can be very lightweight, but it will need to be between 3.5 and 4.0 m
in diameter. Special fixturing on the dome will be required to remove the baffle when changing
to an IR observing configuration, or it might be possible to fold up the baffle in order to remove
it from the light path when it is not in use. Two views of the secondary baffle are shown in
Figure 13.6-23. Presumably the secondary mirror baffle would also be used with any instrument
when working between the UV and the J band.
13.6-26
Figure 13.6-23. Two views of the secondary mirror baffle. The right-hand view is a projection along the optical
axis rather than a perspective view from the position of the focal plane. Since the baffle is behind the secondary
mirror surface, it will appear slightly smaller when viewed from the focal plane.
The combination of the circular baffle behind the secondary and the circular stop at the grating
will function to fully baffle the instrument, but the general problem of scattered light internal to
the spectrograph optics will get some extra help from an additional stop near the primary mirror
vertex. The hole in the central primary mirror will probably be somewhat larger than the
absolute minimum required to provide optical clearance for a 20 arcmin field (this is a subject of
ongoing discussion) but the stop can be just the right size, which is 1.60 m in diameter. At a
distance of 5.5 m from the focal plane, the vertex stop limits the sky illumination incident on the
field lens and aperture plate to the equivalent of an f/3.4 beam. The vertex stop is visible in
Figure 6-11.
13.6.12 GMACS Mechanical Layout
The fly’s-eye configuration for GMACS requires the introduction of the tent mirrors and leads
very naturally to a mechanical configuration which is spread across the diameter of the
instrument platform, but which uses relatively little of the axial depth which has been reserved
for instruments on GMT. This suggests the possibility of mounting the components of GMACS
in such a way that they can co-exist with one or more other instruments mounted below the
instrument platform and which use the remainder of the available volume. Such an arrangement
might lead to greater operational flexibility and at the same time simplify the requirements for
operational support. Achieving a high degree of integration of multiple instruments will require
cooperation between several instrument groups as the instrument concepts and interfaces are
developed. The intention of the GMACS design study at this stage is to make a preliminary
exploration of the possibilities for such an integrated instrument package.
There are six main components of the GMACS instrument package:
13.6-27
(1) The field lens assembly, which contains the last element of the field corrector/ADC. The
field lens is mounted just below the bottom of the instrument platform, and just above the
corrected focal plane/aperture mask.
(2) The aperture mask shuttle, which carries the mask containing multiple entrance apertures
for the imaging spectrographs. The aperture mask must be made of a thin opaque material and
shaped to conform to the curvature of the telescope plane.
(3) The aperture mask changer, which can load one of several aperture masks onto the aperture
mask shuttle. Each aperture mask is custom-made for a particular observation by cutting holes
or slits into a blank mask.
(4) The tent mirror assembly, which contains the diagonal tent mirrors as well as movable
guide cameras for aligning the aperture mask on the sky and comparison sources for active
flexure compensation.
(5, 6) Two optics modules, each of which contains the collimator optics, dichroic mirrors, red
and blue gratings, camera optics and CCD mosaics for two of the four fly’s-eye spectrographs.
In addition to the guide cameras which are part of the focal plane assembly, there will be at least
two image analyzers which are used to maintain active-optics alignment of the primary and
secondary mirrors, and primary mirror figure control. It is not yet clear whether the image
analyzers will be part of the field lens assembly or the tent mirror assembly.
The instrument platform is a relatively thin structure which will need to be reinforced by
additional framework. The mechanical concept for GMACS calls for the 6 main components to
be mounted in a framework which reinforces the instrument platform when GMACS is mounted
on the telescope. The two optics modules are located fairly far from the optical axis and provide
good clearance around the focal area. A pair of rails crossing the instrument platform allow the
field lens and tent mirror assemblies to be moved off to the side of the instrument platform to
make room for other instruments. The same rails are also used by the aperture mask shuttle,
which moves back and forth between the position of the aperture mask changer near the
perimeter of the instrument platform and the on-axis position where the aperture mask is used.
Figure 13.6-24 is an axial view of the six main components of GMACS in the fully-retracted
configuration. The aperture mask shuttle and changer are shown at the top of the diagram. The
field lens assembly and the tent mirror assembly are shown at the bottom of the diagram. The
two optics modules are shown at the left and right. The actual designs of the modules are only
intended to be schematic at this point, but to provide at least some idea of how the various parts
of the spectrograph will fit together.
Figure 13.6-25 shows the components of GMACS moved into position on the optical axis. The
general level of illumination is reduced by many orders of magnitude at the aperture mask, after
which the optical path must be kept completely light tight. The optical baffles which connect the
tent mirror assembly to the optics modules are also only intended to be schematic, but some form
of deployable or removable baffle will certainly be required, as well as a careful seal between the
aperture mask shuttle and the tent mirror assembly. The grating changers are shown
schematically as long linear slides attached to the optics modules, with light-tight covers which
13.6-28
accommodate the maximum extension of the slides.. The alternative of using grating wheels is
also a possibility and both arrangements will be considered in greater detail at a subsequent stage
of the design study.
Figure 13.6-24. Axial view of the six main components of GMACS in the fully-retracted configuration.
13.6-29
Figure 13.6-25. View of the six main components of GMACS configured for use.
“Details” of the field lens assembly and aperture mask changer are shown in Figure 13.6-26.
The field lens assembly moves independently of the tent mirror assembly, so the corrector/ADC
can be used with other instruments, for example with a separate fiber feed for the GMACS optics
modules. The aperture mask changer has a completely stationary storage area for a number
aperture masks, together with a mechanism which can insert or remove a mask from any slot
(including one carried by the aperture mask shuttle), and translate parallel to the optical axis in
order to move from one slot to another.
Figure 13.6-26. Two “detailed” views of GMACS. Left, the field lens assembly. Right, the aperture mask
changer.
13.6-30
The basic structure of the framework which holds the components of GMACS and also
reinforces the instrument platform is shown in the left panel of Figure 13.6-27. The structure
maintains a clear diameter of ~4 m centered on the optical axis when the components of GMACS
are retracted. An axial instrument mounting ring is attached to the bottom of the framework by a
set of jacks which can be used to raise an axial instrument into position along the optical axis, or
lower it to create clearance for the focal plane components of GMACS. The position of the axial
instrument mounting ring needs to be defined with high precision only when it is in the raised
position and engages closely with locating stops on the GMACS structural framework.
Above the axial instrument mounting ring an axial instrument must fit within the 4 m clear
diameter, and also observe the constraints imposed by the location of the two GMACS rails,
which are separated by about 1.5 m and located just above the uncorrected focal plane. Below
the axial instrument mounting ring, the volume available for an axial instrument expands to
essentially the full diameter of the instrument platform. The right panel of Figure 13.6-27 shows
an axial view of the framework mounted on the instrument platform. In this view the main
components of GMACS are shown in the retracted position, and the main vacuum assembly of
NIRMOS (Chapter 13.5) has been inserted along the optical axis.
Figure 13.6-27. Two views of the GMACS structural framework.
NIRMOS and GMACS are shown in two quartering views in Figure 13.6-28. In the left panel
the components of the two spectrographs are shown for clarity without the GMACS structural
framework. In the right panel the components of the two spectrographs are shown in the same
configuration, but with the GMACS structural framework added.
13.6-31
Figure 13.6-28. Two views of GMACS and NIRMOS, with and without the structural framework.
The left panel of Figure 13.6-29 shows NIRMOS in the retracted position and the focal plane
components of GMACS translated into position along the optical axis. An important feature of
the GMACS structural framework is that the flat plane at the bottom, covered with an
appropriate steel grating, provides service access to the components of the spectrograph, as
shown in the right panel of Figure 13.6-29.
Figure 13.6-29. Left, GMACS and NIRMOS in the “GMACS” configuration. Right, GMACS service access.
13.6-32
13.6.13 Summary and Acknowledgements
Although an ambitious baseline concept has been developed for GMACS, with several obvious
descope options, the GMACS design study is far from complete. Extensive work has been done
to develop an optical design, but much more work remains to be done at the conceptual level, for
example on thermal effects and on fabrication and alignment tolerancing. Possibilities for the
mechanical layout have only begun to be explored. The design has proceeded with feedback
from the GMT Science Working Group, and the current optical and mechanical concepts for
GMACS have evolved substantially from the ones which were outlined in the original proposal
written in March 2005. A variety of interface issues with both the telescope and with other
instruments have been identified. Numerous questions call for immediate further study, many of
which are mentioned in the text. The design has progressed to the point where it should be
possible to obtain cost estimates for most of the optical components, but much more work on the
mechanical design will be required before it is possible to make a complete cost estimate for the
instrument.
The GMACS design study has been carried out with the participation of the Australian
instrument groups at Mount Stromlo and the AAT, in particular Peter Conroy, Andrew Granlund,
Ross Zhelem and Will Saunders. Steve Gunnels provided the figures presented in Section
13.6.12, based partly on concepts which were first suggested by Peter Conroy. The VPH grating
calculations presented in Section 13.6.7 were made by Dan Fabricant.
13.6.14 References
Bernstein, R. A., Shectman, S. A., Gunnels, S. M., Mochnacki, S. and Athey, A. E. 2003, in
Instrument Design and Performance for Optical/Infrared Ground-based Telescopes, eds. M. Iye
and A. Moorwood, Proc. SPIE, 4841, 1694
Bigelow, B. C. and Dressler, A. M. 2003, in Instrument Design and Performance for
Optical/Infrared Ground-based Telescopes, eds. M. Iye and A. Moorwood, Proc. SPIE, 4841,
1727
Crampton, D. and Murowinski, R. 2004, in Ground-based Instrumentation for Astronomy, eds.
A. Moorwood and M. Iye, Proc. SPIE, 5492, 181
Dekker, H., D’Odorico, S., Kaufer, A., Delabre, B. and Kotzlowski, H. 2000, in Optical and IR
Telescope Instrumentation and Detectors, eds. M. Iye and A. F. Moorwood, Proc. SPIE 4008,
534
Epps, H. W. 1998, in Optical Astronomical Instrumentation, ed. S. D'Odorico, Proc. SPIE 3355,
111
Epps, H. W. and Sutin, B. M. 2003, in Instrument Design and Performance for Optical/Infrared
Ground-based Telescopes, eds. M. Iye and A. Moorwood, Proc. SPIE, 4841, 612
13.6-33
Fabricant, D. G., Epps, H. W., Brown, W. L., Fata, R. G. and Mueller, M. 2003, in Instrument
Design and Performance for Optical/Infrared Ground-based Telescopes, eds. M. Iye and A.
Moorwood, Proc. SPIE, 4841, 1134
Fabricant, D., Fata, R. G., McLeod, B. A., Szentgyorgyi, A. H., Barberis, J., Bergner, H. W.,
Brown, W. R., Caldwell, N., Conroy, M. A., Eng, R., Epps, H., Furesz, G., Gauron, T. M.,
Geary, J., Goddard, R. E., Hartmann, L., Hertz, E. N., Honsa, M., Mueller, M., Norton, T. J.,
Ordway, M. P., Roll, J. B., Williams, G. G., Freedman-Woods, D. and Zajac, J. M. 2004, in
Ground-based Instrumentation for Astronomy, eds. A. Moorwood and M. Iye, Proc. SPIE, 5492,
767
Glazebrook, K. and Bland-Hawthorn, J. 2001, Pub. A. S. P. 113, 197
13.6-34