The primary scientific goal for use of a deployable integral field spectrograph is the tomography of galaxies and pre-galactic fragments.
The goals are to determine the kinematics for the gas and stellar components of distant galaxies and the ensembles of stars that make up the pre-galactic fragments in the early universe, and to quantify the star formation rate and chemical composition of those objects.
In order to achieve this, spectroscopy of the HII complexes and underlying stars making up these objects is required for the measurement of velocities and abundances. As discussed in Chapter 2, spectral resolving powers of R = 5000 to 10,000 are desired to provide adequate velocity resolution, and also to limit the contamination of the terrestrial night sky OH emission. A wide field of view (FOV) is desired to observe a statistically significant sampling of targets within a reasonable amount of telescope time, and to determine environmental variations across the observed FOV by sampling many objects within that field.
The science requirements are:
- Operational spectral window of 1-2.5 microns
- Image quality of 0.05 to 0.10 arcseconds
- FOV of 1.5-2.0 arcminutes in diameter
- IFU (integral field unit) FOV of about 1 arcsecond diameter
- IFU sampling of about 0.1 arcseconds and at least one deployable IFU with ~ 0.005-0.01 arcsecond sampling to exploit the full diffraction limit of GSMT for high surface brightness scenes
- Ability to observe greater than 10 targets per observation
- Spectral resolving powers of up to 5000 and 10,000
- Spectral coverage per exposure of complete J-band (1.14-1.36 microns), H-band (1.40- 1.82 microns), or K-band (1.92-2.58 microns) at R ~ 5000
The following instrument parameters were selected:
- Utilization of the multi-conjugate adaptive optics (MCAO) image surface to make use of the good image quality
- 100 mm beam diameter in the spectrograph
- Two IFUs per spectrograph
- 2 K detector format
- Goal of 26 IFUs implying the need for 13 spectrographs
- IFU FOV of 1.5 x 1.5 arcseconds with ~ 0.05 arcsecond slices, giving a total of 31 slices per IFU and 0.15 x 0.15 arcsecond fields for the high resolution IFU arms (the high resolution arms were not specifically designed in the present study)
- Deployable relay optics converting the MCAO f/38 beam into an f/128 beam at the IFU slicer input
- Telecentric input and output optics for the deployable relay and IFU
- Cold stop within the deployable relay arm
- IFU relay produces f/11.5 input for spectrograph
- f/4.7 spectrograph camera
The NIRDIF (near-infrared deployable integral field spectrograph) concept is largely based upon a concept originally developed by the Astronomy Technology Centre in Edinburgh, Scotland, as a concept for the Gemini infrared multi-object spectrograph (GIRMOS).1 The major change in the implementation of NIRDIF as compared to GIRMOS is the exclusion of the anamorphic image relay within the NIRDIF design. We chose to eliminate that element owing to the complexity of the mirror surfaces; by following the GIRMOS approach, these mirrors would have to be toroidal and would likely be quite difficult to align. Such complexity may be a significant cost and risk factor in an instrument that contains as many potential spectrographs as are considered for NIRDIF. The disadvantage is the need for a factor of two additional spectrographs, because only two IFU units can feed into a single spectrograph, whereas the anamorphic relay system would allow four IFUs to feed into a single spectrograph. The other disadvantage is that the NIRDIF concept results in a factor of two difference in "true" spatial sampling for both spatial dimensions. Our view is that the IR detector pixels could, in theory, be co-added in the spatial dimension (orthogonal to the spectral) to provide equivalent spatial sample in the two-dimensional spatial domain. However, a future study should be conducted to evaluate the risk and cost trades for inclusion of the anamorphic relay system.
The following assemblies and subassemblies are included in the NIRDIF design:
- Deployable image relay assembly
- Image relay optics
- Pupil cold stop
- Image slicer (IFU) assembly
- Calibration assembly
- Cryogenic housing assembly
The current design effort only studies items A, B, and C. Future studies will be required to understand how to calibrate such an instrument and how to house it in an appropriate cryogenic chamber.
The overall instrument layout for a single IFU and spectrograph is shown in Figure 1. Two views are shown: the top shows one IFU feeding the spectrograph, and the bottom shows the other IFU feed into the same spectrograph. Note that they feed onto opposite sides of the detector. Figure 2 displays some representative spot diagrams.
The study produced by the ATC group describes an imaginative technique for accessing the targets scattered about the image field with a deployable arm that contains the image relay optics. We utilize that same concept for the NIRDIF instrument. Those optics are identified in Figure 1.
The input to the relay is assumed to be telecentric. The output of the relay has also been set to be telecentric so that the unit can be designed independently of the slicer and spectrograph.
A 1.5 x 1.5 arcsecond FOV is relayed from the f/38 output of the MCAO feed to the entrance of the image slicer where the focal ratio has been increased to f/128. This magnification is desired in order to allow the mirrors in the IFU assembly to have a reasonable thickness of about 1 mm.
A pupil image is created within the relay so that a cold stop can be placed to filter out background thermal radiation prior to image slicing or spectral dispersing. This requires the full deployable arm to be contained within a cryogenic environment operating at a temperature near that of liquid nitrogen. It is most likely that the full set of deployable arms will be contained within a single, but large, cryostat with a large optical window. Further design effort will be required to study the available options.
The optics of the relay are not seen as a risk. Two doublet lenses are utilized out of SF6 and BaF2 glass. The first lens is about 12 mm in diameter; the second has a diameter of about 34 mm. Although BaF2 may be a difficult material to obtain in large sizes, at these modest dimensions, adequate BaF2 should be available to make the requisite number of lenses for all of the deployable probes.
The deployable arm folds about the optical axis of two folding mirrors. Figure 1 shows two such mirrors that could form one rotation axis for deployment. A second, similar set of mirrors will also be required (but were not included in the optical design) for the second rotation axis that allows the deployable arm to function similar to a human arm with rotation about a shoulder and elbow. Such movement allows the probes to cover a significant area and range of motion.
The slicer selected is that proposed by the ATC in their GIRMOS study. Two sets of sliced mirrors are used to slice and stack the image. The input and output of the slicer is telecentric, allowing the slicer to be designed and built as a completely independent entity from the relay optics and spectrograph. The slice also relays the focal ratio from the f/128 input to f/11.5. See section 220.127.116.11 for a comparative exploration of different image slicing options.
The spectrograph is an all transmissive instrument with classical reflection gratings. The f/11.5 collimator takes the light from the two IFUs feeding it and produces a 100-mm beam. The spectra are imaged by an f/5 camera onto a 4 K HgCdTe detector.
A field lens exists at the entrance of the spectrograph and is made from fused silica. The collimator is comprised of SF6 and BAF2 lenses with spherical surfaces.
The camera contains a total of four spherical lenses made from the following glasses: BAF2, SF6, CaF2, and fused silica.
None of these optical elements is considered high risk, other than the fact that several spectrographs will be fabricated, requiring a large number of identical elements.
We have not yet evaluated the efficiency of this instrument.
No estimate for cost has yet been made.
The major technical risks lie in the following items:
- Development of a cryogenically cooled, deployable relay
- Development of the image slicers
- Packaging of the full set of spectrographs
Other components are not considered high risk, except for the MCAO feed and those mentioned explicitly as risk items.
The next level of GSMT effort will fund further studies of these items, the tradeoff in the use of the anamorphic mirrors versus the number of required spectrographs, and a further design effort in the development of the high resolution arms.
Short-term thinking on this issue has been constrained by the requirements which were set forth at the top of this document: operating range 1-2.5 microns, 1 x 1 arcsec IFU with 0.1 arcsec sampling, resolutions 1000-5000, up to 50 deployable IFUs. This may meet the demands of programs surveying distant galaxies (where the relatively coarse sampling is a consequence of the surface brightness of these objects), but may not adequately meet the requirements of research programs that want to sample near the diffraction limit.
One could envision a WF/PC-type instrument containing both the relatively coarse as well as near-diffraction-limited IFU/spectrograph assemblies. Extending the wavelength coverage beyond 2.5 microns would have serious implications for the design of the spectrograph section.
This instrument will require a significant number of infrared arrays. Assuming that one will be able to multiplex at least two IFUs (and perhaps 4-5) into a spectrograph module, 10-25 large format arrays will be required. This will represent a substantial expense that may be mitigated by the implementation of the anamorphic pre-optics currently left out of the design as described earlier.
A more serious issue would be the implications of extending the wavelength coverage of the instrument beyond 2.5 microns. By the time the detector specifications for this instrument are completely defined, it is likely that large format detectors may exist in both HgCdTe and InSb. The developments in both detector formats should be monitored. For example, one route to large format HgCdTe arrays would replicate the detector on a silicon, rather than sapphire substrate, as this would provide a much more homogeneous behavior under thermal cycling. However, for backside illuminated arrays, this would seriously compromise the spectral response even in the short J-band, and preclude operation short of 1.1 microns. Although this would leave an uncomfortable spectral gap between the CCD and IR coverage, it could simplify the NIRDIF instrument design in terms of glass choices and chromatic performance requirements.
As will be discussed later (section 4.7.4), the MCAO imager is a costly and daunting instrument if it is desired to fully sample the entire MCAO FOV. An alternative approach is to utilize a set of "deployable" imagers to sample the full FOV, but with significantly less imaged field.
The deployable image relay arms of NIRDIF could potentially serve as relays for small fields of view to imagers. The NIRDIF image slicer and spectrograph could be pulled out of service and replaced by a relatively simple imager with a 4 K detector. The imager would have optics that relay the f/128 beam back to f/38. This would give an imaged field of about 16 arcseconds on a side. If two relays feed in to the same imager, each could image an 8 x 16 arcsecond image onto the detector. The relay optics will need to be considerably larger than currently designed. It is more likely that the imaged area will be considerably smaller at the 3-4 arcsecond size.
The following is a comparative study of image slicing options for an instrument such as the near- infrared deployable integral field spectrograph (NIRDIF).
The broad parameters for studying the options are:
|Telescope:||GSMT 30-m aperture|
|Focal Ratio:||f/15 nominally; ~f/30 at output of MCAO|
|Plate Scale:||0.23 arcsec/mm at f/30|
|IFU FOV:||1 x 1 arcsec (nominal)|
|Resolution:||0.1 arcsec (total of ~ 100 elements)|
|Spectral Res:||1000, 5000|
|Number of IFU||~50|
Three basic concepts of integral field spectroscopy are summarized by Allington-Smith et.al.2 The first uses an array of lenslets at the telescope focal plane to form an array of pupil images onto a mask at the input of the spectrograph. Each of these images produces a spectrum, although the detector array must be rotated to avoid overlap. The second concept also uses a lenslet array at the telescope focal plane to image the pupil onto optical fibers, which can be arranged at the other end into a linear array (or several) at the input of the spectrograph. The third concept, the image slicer, achieves a similar repackaging of a square/rectangular field onto a slit utilizing reflective optics.
In searching through the recent literature, primarily SPIE proceedings and often outdated Web sites, no examples of the first type of integral field unit (IFU) for use at infrared wavelengths in existing instruments were found. Aside from the need to precisely align the array to staircase adjacent spectra without overlap, this design suffers from the problem that slit mask spectrographs face: the loss of either spectral coverage or spectral resolution due to the two- dimensional input to the spectrograph.
Such an instrument (OH suppressing infrared imaging spectrograph (OSIRIS)) has been proposed by the UCLA group (PI James Larkin) for use at the Keck adaptive optics (AO) station. This instrument will use a precision lenslet array to form pupil images at each spatial point in the field, which are then fed into the spectrograph. The array is physically rotated to give two pixels between adjacent spectra. With a 2K x 2K HgCdTe array, this configuration would permit either a 16 x 64 lenslet array with broadband (z, J, H, or K) coverage, or a 64 x 64 lenslet array with narrowband filters. A decker would select between the two. Because adjacent spectra are offset in wavelength, there is some concern that the proposed spacing of two pixels would be insufficient to deblend them, particularly because there will be strong OH sky lines in each spectrum. The deblending operation, if it is possible, will be very software intensive.
There is also the effort to build a similar type of instrument in the optical called MEIFU (million element integral field unit). Further discussion on this instrument is given in Section 4.7.5.
A number of fiber-fed IFU spectrographs exist in the concept, construction, and working phases, so the technology, at least at visible wavelengths, is reasonably mature. The use of micro-lens arrays at the input and output of the fibers can result in almost complete coverage of the input focal plane and spectrometer slit fields (the fibers themselves have significant gaps between the light-conducting cores).
Two concerns of relevance that this approach has to the NIRDIF instrument are operation at wavelengths to 2.5 microns and at cryogenic temperatures, which is required for acceptable performance in the thermal part of the K-band. Risk elements are the transmission of the fiber material and the durability of the bond between the micro-lens array and the fibers over cryogenic cycling. Zirconium fluoride has been used for infrared fiber applications, as it has good transmission, even into the 3-micron region. However, this material is prone to developing micro- cracks, particularly where the fibers bend, which can result in decreased efficiency. The team building SPIFFI (spectrometer for infrared fiber-fed field Imaging) plans to use OH-free silica fibers with doped silica cladding and polyimide buffering, claiming that the transmission in the K- band is good over the short fiber length. They eliminate the need for a separate micro-lens array by flaring the end of each fiber and polishing it to the required figure, creating the micro-lens out of the fiber itself.
Image slicing IFUs use a stack of thin rectangular mirrors to slice the focal plane in one dimension and a second set of mirrors to rearrange the slices along the input slit to the spectrograph. The "3-D" spectrograph constructed at MPI is a functional near-IR IFU spectrograph that rearranges a 16 x 16 element focal plane onto a 256 x 256 NICMOS (near- infrared camera and multi-object spectrometer) HgCdTe array. There are some compromises: the 3-D image slicer is at ambient temperature, although the spectrograph is cryogenic, and the two- mirror system cannot align the slices in the dispersion direction, resulting in a "staircase" arrangement of the reimaged focal plane. A much more complex IFU under construction for the GNIRS (Gemini near-infrared spectrograph) instrument should give a linear rearrangement of the slices, but the challenge of actually fabricating the unit has already led to the abandonment of the higher spatial resolution (0.05 arcsec) IFU.
A major risk area for the image slicer is that of fabrication and alignment. The 3-D instrument utilizes flat mirrors for both arrays. This results in a slight defocus at the ends of each linear slicing element. For a small (16 x 16 for 3-D) array, this is apparently not too much of a problem, and would probably not be serious for a similarly small format IFU for the GSMT instrument. One feature of this design is that the reimaged slit segments do not lie in a plane, which must be accommodated by the spectrograph design. The near-infrared integral-field spectrograph (NIFS) utilizes three arrays of spherical diamond-turned mirrors. The slicing array at the focal plane reimages the pupil onto an array of pupil mirrors that then reimage the reformatted focal plane onto a field mirror array that feeds the spectrograph collimator at a focal ratio appropriate to the spectrograph design. This adds a level of complexity to the fabrication and alignment, but does permit one to do the slicing at very slow focal ratios with manageable sized optics, and then convert the reassembled focal plane to a faster focal ratio to keep the spectrograph size reasonable.
NOAO is collaborating with STScI and NASA Goddard on a near-IR spectrograph to evaluate a digital micro-mirror array at the input focal plane. STScI views this as a concept evaluation for a proposal to construct a multi-object IR spectrograph for JWST (James Webb Space Telescope). From a science perspective, this is an alternative concept to deployable fibers or slit masks for spectroscopy of multiple objects in a field. But as an IFU, it seems to offer no advantages and many disadvantages over any of the three concepts discussed above, so it received no further consideration. The development of cryogenic micro-mirror arrays may be technically daunting. In addition, the digital micro-mirror design results in a reimaged focal plane that is not normal to the optical axis, requiring exotic optics in the spectrograph itself. An alternative concept of digital micro-shutters, which was recently selected as the favored concept by JWST, is much more straightforward optically but faces the same technical challenge of reliable cryogenic operation. Both of these micro-mechanisms vignette the input beam to some extent.
The image slicer concept appears to be the desired choice for the NIRDIF instrument. A brief listing of reasoning follows:
- Throughput. Being reflective, an image slicer has potentially very high throughput as well as achromatic behavior throughout the 1-2.5 micron region.
- Diffraction. At the proposed IFU sampling of 0.1 arcsec, diffraction losses are not serious. (At 2.2 microns, the diameter of the Airy disk is 0.036 arcsec.) The image slicer retains the full resolving power of the telescope along the slice direction, whereas a micro-lens array is limited by diffraction in both axes.
- Reliability. This is hard to quantify. However, the loss of a fiber or two, particularly near the center of an IFU array, would compromise the usefulness of that IFU type. A fiber multi-IFU instrument would have to be designed to permit the replacement of IFU modules that fail.
- Fabrication. This is a thorny issue with implications for both concepts. A number of fiber-fed IFU systems exist or are in fabrication. Aside from 3-D, there may be one other IR image slicing IFU around. The 3-D image slicer is not cryogenic. NIFS will be an interesting test case, and the instrument team at the Australian National University (ANU) is progressing well at least on the mechanical front, because it recently successfully cooled down the cryomechanical system. A fiber-fed IFU requires a large number of individual fibers to be assembled into a fixture that provides and maintains the alignment. The image slicer IFU uses arrays of small mirrors that must be fabricated (typically using diamond-turning technology), assembled, and aligned to an extremely high degree of precision. In addition, the assembly must maintain its alignment when cooled; this would require that the mirror arrays and the assembly fixture be made of the same material. On the other hand, an instrument that may contain as many as 50 IFUs may be able to take advantage of replication technology in generating copies of the image slicer (one could fabricate and align one unit and use that as a template to replicate the other units). Each fiber IFU, on the other hand, would require individual assembly of the fibers.
- Feeding the Spectrograph. The strawman spectrograph concept for NIRDIF has the IFU(s) as part of the spectrograph module, fed by roving optical relay arms that would move about the science field. The requirement to cover as much spectrum as possible, particularly at a resolution ~ 5000, dictates the largest array format likely to be obtainable, probably 4K. The relatively small number of elements per IFU will utilize only a small portion of the spatial dimension of the array, so multiplexing as many IFU inputs into each spectrograph will make the most efficient use of each array and limit the number of spectrograph modules (and arrays, electronics, etc.). One might expect this to be accomplished more efficiently with fibers, but some ingenious thinking will likely devise a method to feed multiple image slicers with cleverly designed relay arms (as is done in the GIRMOS design).
- GIRMOS Development Studies Technical Report C for Gemini Contract #01039, PPARC (2001)
- Allington-Smith, J., Content, R., and Haynes, R. "New developments in integral field spectroscopy". Proc. SPIE 3355, 196 (1998).