Chapter 4

Chapter 4, The Point Design

Section 4.7.1: Wide-Field Multi-Object Multi-Fiber Optical Spectrograph (MOMFOS)


Two major science programs drive the need for wide-field multi-object spectroscopy that exploits the increased light gathering power of GSMT: tomography of the universe and studies of the stellar populations comprising nearby galaxies—a key step toward understanding how galaxies are assembled. A practical driver is the need for an instrument that can exploit the power of GSMT during those times (at least 20% and perhaps more) when high Strehl adaptive optics (AO) imaging and spectroscopy is precluded. Tomography of the universe

The goal is to map out the large scale structure of the universe at redshifts greater than 3 and to link the emerging distribution of gas and galaxies to the Cosmic Microwave Background.

To map large scale structure using both galaxies and gas as probes requires (1) spectroscopy of 106 galaxies at a resolving power of R ~ 2000, covering a wavelength regime from about 500 to 1000 nm, and (2) spectroscopy of 105 quasars at a resolving power of about 15000. The spectral regime for this latter study would push towards the blue with a desirable blue end of about 350 nm. Targets will be fainter than 25 AB magnitude and could be as faint as 27 AB magnitude. Their density on the sky can be as high as 200,000 targets per square degree, with a typical size ranging from a few tenths to one half of an arcsecond in diameter.

The ability to observe samples of this size in a reasonable period of time requires simultaneous observation of about 1000 objects over a field of view (FOV) no smaller than 20 arcminutes in diameter. Design of a multi-object spectrograph of this power would enable completion of the large scale structure surveys in approximately three years with GSMT. Such a study utilizing current instruments on the Gemini telescopes would take about a hundred times longer to accomplish. Study of Galaxy Clusters

The goal is to map out kinematic motions and the chemical enrichment of the streams of intracluster stars that trace the history of galaxy mergers and interactions in nearby galaxy clusters.

A large, seeing-limited FOV with optical and near-IR multi-object capability is required to densely map out the streams with a spectral resolving power of greater than R~2000. Full spectral coverage is not required. The desired wavelength regime would be redward of about 500 nm. INSTRUMENT REQUIREMENTS

The science requirements are:

  1. Operational spectral window of 350 to 1100 nm for the optical, and a goal of J- and H- band coverage in the near-infrared
  2. A seeing-limited FOV of at least 20 arcminutes in diameter
  3. Image quality of better than 0.5"
  4. Ability to observe up to about 1000 targets per observation
  5. Resolving powers of 1000, 5000, and up to ~ 20,000
  6. Spectral coverage per exposure of:
    1. ~ 1 octave at R ~ 1000
    2. ~ 0.2 octave at R ~ 5000
    3. ~ 50 nm at R ~ 20000
  7. Limiting magnitude of R ~ 26.5 at R ~ 1000

The science requirements, the telescope point design under consideration, and the sheer size and detector real estate covered by a Cassegrain implementation of a wide-field MOS have led us to explore a prime focus implementation of the multi-object spectrograph with the following requirements:

  1. Parks and Barden1 showed that the diameter of a circular entrance aperture of ~ 1.4 times the FWHM (full width half-maximum) delivered image is needed to give optimal signal-to-noise throughput for sky-limited targets. This results in 0.7 arcsecond apertures for 0.5 arcsecond seeing.
  2. Telescope aperture, entrance aperture, and resolution requirements imply a minimum monochromatic beam diameter of 400 mm.
  3. A prime focus implementation implies the use of fiber optics feeding to spectrographs located off the telescope. This results in a fiber cable run of between 50 and 60 m in length.
  4. Ability to compensate for wind-buffeting of the telescope with an adaptive mirror as part of the prime focus corrector assembly.
  5. Use of fibers requires transfer of f/1 prime focus beam to between f/4 and f/5 for optimal fiber illumination, resulting in minimal focal ratio degradation and effectively complete image scrambling.
  6. A fiber size of 500 microns subtends 0.7 arcseconds on the sky in an f/5 beam.
  7. Spectrograph collimator selected to be f/4.5 to account for residual focal ratio degradation and to ensure maximal collection of light from the fibers.
  8. Spectrograph camera goal to image 500 micron fibers onto 5 pixels at f/0.69, requirement to image onto 6 pixels at f/0.85.
  9. Spectrograph stability requiring less than 1/10 of a pixel drift in the detector over the course of 1 hour, and of less than 0.75 arcminutes of rotation in the detector over the course of 1 hour.
  10. Detector charge shuffle mode with on-sky beam switching required to achieve the required level of sky subtraction (up to a part in a few thousand).
  11. Nod-shuffle implies fiber separation on the detector of greater than 2.5 times the FWHM image of the fiber (assuming a worse case of the fiber being imaged as a Gaussian profile). If the fiber is imaged onto 6 pixels (f/0.85 camera), the minimum separation required is 18 pixels, allowing about 230 fibers per 4K detector.
  12. Assuming a 4K detector format for the spectrograph with an internal focus, three such spectrographs would be required to allow the use of 700 fiber probes in total.

The choice to explore a prime focus MOS is driven primarily by the desire to shrink the physical size of the field by using a fast focal ratio. With an f/1 focal ratio, a 20 arc-minute field of view is only about 175 mm in diameter, compared to 2.6 m for an f/15 focal ratio. Although the fast focal ratio requires the use of micro-lenses on the ends of the fibers, the relatively small diameter of the field allows the use of current technology atmospheric dispersion compensation and image correction with reasonably sized optical elements. The disadvantage of the prime focus implementation is the requirement to use fiber optics rather than slits. However, this trade may be mitigated through the use of nod-shuffle mode observing, in which fibers will be able to achieve very high precision sky subtraction.2 How much improvement can be gained through nod-shuffle with fibers is not yet clear, and is identified as one of the areas requiring further exploration at the end of this section. INSTRUMENT DESCRIPTION

The multi-object multi-fiber optical spectrograph (MOMFOS) consists of the following assemblies and subassemblies:

  1. Prime focus assembly
    1. Atmospheric dispersion compensator
    2. Image relay mirrors
      1. One mirror with adaptive actuators
      2. Second mirror with tip-tilt actuators
    3. Correcting lenses
  2. Fiber positioner assembly
    1. Echidna-type fiber positioner
    2. WFS probes
    3. Acquisition imager
    4. Fiber micro-relay lenses
    5. Fiber optics
    6. Fiber cable
    7. Field rotator
  3. Spectrograph
    1. Fiber interface
    2. Field lens
    3. Order separation filter
    4. Collimating mirror
    5. Articulation arm
    6. Grating and grating mount
    7. Camera corrector lenses
    8. Camera mirror
    9. Field flattener lens
    10. CCD array
    11. Cryostat
  4. Calibration Unit
    1. Projection screen
    2. Wavelength calibration lamps
    3. Flat field lamps

A schematic overview of the instrument is shown in Figure 1.

Figure 1 Schematic view of MOMFOS on the point design GSMT. Prime Focus Assembly

Figure 2 Detailed view of MOMFOS prime focus assembly.

The prime focus assembly is shown in Figure 2. Light from the primary mirror first passes through a set of risley prisms for atmospheric dispersion compensation (ADC). An image is formed just in front of the first set of correcting elements. A set of relay optics (the two mirrors of the image relay) reimages the image surface at the secondary location indicated in the figure. The two sets of correcting lenses minimize aberrations from the primary mirror.

The calibration unit mounts in the space between the ADC unit and the initial image surface. The wavefront sensors (WFSs), fiber probes, and acquisition imager all interface at the fully corrected, secondary image surface.

Atmospheric Dispersion Compensator

The current design allows excellent correction over the 20 arcminute field up to a zenith distance of 60 degrees. Images across the FOV retain a 90% encircled energy diameter of 120 microns, or 0.83 arcseconds across a bandpass from 340 to 1024 nm at 60 degrees zenith distance. The 50% encircled energy diameter is contained within a 0.5 arcsecond image. At 30 degrees zenith distance, image quality is significantly better with the 90% encircled energy falling within a 0.41 arcsecond image diameter across the full field and wavelength band. Figure 3 shows the encircled energy performance of the ADC, corrector, and relay mirrors.

Figure 3 Encircled energy diagrams for the prime focus corrector and ADC assembly. Top figure
is for Zenith pointing. Middle figure shows image quality for 30-degree Zenith distance. 60-degree
Zenith distance performance is displayed in lower figure. Each panel shows the performance as a
function of field. Plots are for the polychromatic image that covers the wavelength range of 340
to 1024 nm. The fiber aperture of 0.7 arcseconds corresponds to a radius of 50 microns in these plots.

The optics for the ADC prisms are made from BK7 and LLF6 with dimensions of about 700 mm diameter, 40 mm thick, and with wedge angles of 2.5 to 3 degrees. The BK7 glass should be relatively straightforward to obtain. The LLF6 material may prove more problematic. This glass has previously been put into ADC units (Blanco wide-field Cassegrain corrector) of comparable size, but the availability from a vendor is spotty. Schott may require a custom production run to make the quantity of LLF6 glass required.

The two sets of prism assemblies will each require independent rotation via a rotary stage assembly. Range of rotation for each prism set is 180 degrees.

Image Relay Mirrors

Wind-buffeting on a 30-m telescope is a significant source of image degradation and must be corrected with a deformable adaptive mirror (see Section 5.5, Characterization of Wind Loading). A prime focus, by definition, doesn't have any mirrors that are acceptable for this purpose, other than the primary mirror itself. Either a fold mirror or mirror relay must be implemented. A fold mirror was rejected due to the large size required and because of the non- asymmetrical nature of the resultant fold. Instead, we opted for a relay of the image by way of a two-mirror system similar to an image relay. Figure 2 shows the optical path of the relay.

This design has several advantages: one or both mirrors can be made into an adaptive mirror to compensate for wind-buffeting; the mirror environment can be environmentally sealed, minimizing dust contamination to the adaptive mirror components and mirror surfaces; internal baffles can (and must) be implemented, resulting in suppression of extraneous sky light; and the figure of the mirrors can be utilized to help correct the images across the FOV. Drawbacks to this design include that (1) the mirrors are still rather large in size, with diameters of about 2 m each, and (2) central obstruction losses are not insignificant (about 10 to 15%).

The current design study places an image of the primary mirror on the first image relay mirror. This mirror will be the adaptive mirror for wind compensation. The second image relay mirror would likely have a tip-tilt mechanism to allow for on-target and off-target beam switching (as described later) without requiring full telescope nodding. Struts for the relay central baffle are aligned with the telescope struts for the prime focus support to minimize obstruction losses and diffraction spikes.

In addition to correcting for wind-buffeting, the adaptive mirror will also correct boundary layer turbulence that will be indistinguishable from wind-buffeting. As discussed in Appendix 4.6.A, slightly improved optical imaging will result. It is anticipated that median seeing can be reduced from 0.7 arcseconds to 0.5 arcseconds FWHM, with a modest amount of correction (about 1000 actuators) and native guide stars. Further design effort must be made to determine (1) how many guide stars are actually required to achieve 0.5 arcsecond median images and (2) whether laser guide stars could be utilized.

It may be possible to further enhance the image performance if the second mirror in the system is made adaptive as well. Unfortunately, the current design does not place the conjugate of the second mirror at the desired height of a few kilometers above the primary. The second image relay mirror is therefore not envisioned to be fully adaptive at this time.

Correcting Lenses

The relay mirrors are insufficient by themselves to fully correct the aberrations of the prime focus image. Two sets of lenses have been added to provide excellent image quality at the final image surface. The first set is positioned near the initial focus; the second set is just before the final focus. The first set of lenses also places the pupil onto the first of the two image relay mirrors.

Glass types are fused silica, LLF1, and UBK7. The largest element, the LLF1 lens, is about 500 mm in diameter. The LLF1 lens and one of the UBK7 lenses are dome shaped, requiring slumping or significant glass removal, but are not seen to be significantly risky.

The resultant image surface produced by the corrector is flat. The telecentric angle at the edge of the field is 5.2 degrees. The choice of fiber positioner (described below) allows for proper alignment of the fibers on the focal surface and alignment with the telecentric angle.

Image quality diagrams (encircled energy) for the image relay, correcting lenses, and ADC are shown in Figure 3. Fiber Positioner Assembly

Figure 4 Close-up schematic of the fiber positioner concept based upon the Echidna positioner
developed by the Anglo-Australian Observatory.

Figure 4 shows a close-up schematic of the fiber positioner assembly and related hardware.

Fiber Positioner

The positioner concept for the Subaru telescope instrument, FMOS, appears to be well-suited for use on the GSMT point design telescope. That concept, called the Echidna, allows the fibers to tip-tilt about a spherical pivot to acquire the target object. Each fiber protrudes some distance above the pivot, moving within a restricted FOV as the probe is tilted. If the angular tilts are kept small, the degradation of the input focal ratio will be held to a minimum.

The Echidna instrument being developed by the Anglo-Australian Observatory (AAO) contains 400 fiber probes to cover an FOV of 30 arcminutes at the prime focus of the Subaru telescope. The probes are moved by sawtooth vibration induced by piezo cylinders, causing the probes to tip-tilt about their pivots. The present AAO/Subaru concept does not have any encoding to determine the angular location of the probe. Positioning is conducted in an iterative manner using a TV camera to determine the location of the probes after individual iterations. Positioning uncertainty is about 10%, with three iterations being adequate for reaching the desired location. Positioning time for the 400 fibers of the Echidna instrument is expected to take about five minutes in total.

The physical scale of the Echidna instrument is also a very good match for the MOMFOS positioner. The probes are spaced at 7 mm spacing for the Echidna. 6 mm spacing is envisioned for the MOMFOS instrument. This would allow the placement of about 700 fibers on the GSMT. Configuration time would then be about a factor of two larger than that for the Subaru Echidna, or about 10 minutes.

One significant advantage of the Echidna approach over the placement of fibers onto a flat or curved surface is the ability to curve the pivot surface and utilize different probe lengths to allow the fiber probes to simultaneously match the focal surface of the telescope, while retaining proper alignment of the fiber optical axis with the telecentric angle. The drawback is that individual fibers can only reach a limited area on the sky, and the tip-tilt angle required to position the fiber will cause a slight deviation in alignment between the fiber and the telescope.

It would be highly desirable to develop an encoding mechanism on the fiber probes in order to eliminate the need for the TV viewing system and to allow for more rapid positioning of the probes. Encoding might also allow the possibility to "tweak" the fiber positions and compensate for atmospheric refraction across the FOV as the telescope tracks across the sky. However, the current technique for iterative positioning will work for the MOMFOS instrument.

Wavefront Sensor Probes

As mentioned previously, wind-buffeting on the telescope will require the use of an adaptive mirror in the optical train feeding the native seeing wide field. That adaptive mirror will be one of the two image relay mirrors discussed previously. In order to determine the correction needed, WFSs located in the prime focus field are required.

Six to 10 sensor probes will be located about the circumference of the 20 arcminute science field. Each probe will be capable of being moved in radius and angle ranging from field center to a zone outside the FOV. It is hoped that the number of guide stars required will allow the use of a zone surrounding the science field in order to minimize obstruction of target objects by the probes. Further design effort will be required to refine the implementation of the WFS arrangement.

The specific design for the WFSs has not been completed. They are envisioned to be Hartman- type sensors that would sample the wavefront on a 30 by 30 grid.

It is anticipated that the WFSs will also serve as the closed-loop guiders for telescope tracking.

Acquisition Imager

An imager is required to point and align the telescope onto the desired field of target objects. Due to lack of space between the final lens of the corrector and the image surface, the fiber probes will have to be retracted in order to insert such an imager.

Once the imager acquires the target field, one or more of the WFSs will be brought into position to "lock" onto the target field by initiating closed-loop guiding. The imager will then be retracted and the fibers moved back in to position.

Fiber positioning will take place while the fibers are retracted and the telescope is acquiring the target field. The camera needed to position the fibers will be mounted on the back side of the acquisition imager, but will have the added ability to raster across the fibers, as is being done for the Echidna positioner on Subaru.

It will be critical that the fibers are not subjected to any forces during reinsertion so that they remain in place on their targets. The static friction of the Echidna probes should be adequate to prevent movement of the fibers while the assembly is translated in and out of the image surface.

Fiber Micro-Relay Lenses

The final focal ratio of the point design GSMT prime focus is f/1, which is too fast for acceptance by the fiber optics. A relay lens is required to slow the beam down to a reasonable value for the fibers. The numerical aperture of typical multi-mode fibers is about f/2 to f/2.5. Fibers have nearly ideal preservation of the focal ratio at these fast f/ numbers. However, multi-mode fibers fed near their numerical aperture tend to retain imaging information of the radial distribution of light on the input end of the fiber. This means that if a star wanders about the input face of the fiber, the output image will modulate in FWHM, depending on the radial position of the star. At f/ numbers slower than f/4 to f/5, the radial information effectively becomes perfectly and completely scrambled. But, in order to prevent significant focal ratio degradation, the f/ number should not be slower than about f/4 to f/5. Consequently, we desire to feed the fibers on MOMFOS at a focal ratio of somewhere between f/4.5 and f/5. Figure 5 displays the focal ratio degradation (FRD) and radial scrambling performance for typical fibers.3

A micro-relay lens is required on the tip of each fiber. Figure 6 shows the design for a two-lens concept that actually images a quasi-pupil onto the end of the fiber. The light is contained within an f/3 cone and 100% of the light is coupled into the 500-micron fiber as shown in Figure 7.

Figure 5 Plots showing typical focal ratio degradation and radial scrambling performance for fiber
optics as a function of input focal ratio.

Figure 6 Micro-lens design for fiber ends. The light comes to a focus (on the right side) and a
quasi-pupil is imaged onto the fiber (left side). A field stop is required at the image location
(shown by the red lines). Figure 7 Footprint diagram for the light imaged by the micro-lens relay onto the end of the 500-
micron fiber.

For optimal imaging performance, the lens should be axially aligned with the chief ray. This is also true of the fiber; the lens alignment on the fiber is straightforward and identical for each fiber regardless of the field position. The pivot plate of the fiber positioner will be fabricated to properly align both fiber and lens with the telescope.

Attempts were made to image the star itself onto the end of the fiber, but aberrations prevented good light coupling to the fiber. Imaging the pupil results in much better coupling efficiency, but will require the addition of a micro-lens assembly on the output end of the fiber. The design for that lens has not yet been carried out. It should have roughly the same form as the input end, but will be designed to relay the light into an f/4.5 beam rather than f/1.

Fiber Optics

A typical silica fiber might have the dimensions of a 500-micron core surrounded by a 25-micron- thick silica cladding and polyamide buffer, giving a total diameter of 590 to 600 microns.

The fibers would need a length of about 60 m in order to transmit the light from the prime focus location down to an environmentally stabilized spectrograph chamber located near one of the Nasmyth platforms of the telescope.

The full physical diameter of the 20 arcminute field, at f/1, is 174 mm. With a 6-mm spacing between fiber probes, a total number of about 700 fibers could be positioned within the field. A total of about 42 km of fiber is required!

The type of fiber would preferentially be a silica material with good IR performance and good optical-to-UV transmission. The STU fiber material produced by Heraeus Amersil would likely be the most desirable. A standard low OH fiber would be the desired second choice, but at the cost of relatively poor UV performance. High-OH fiber should be avoided, as the OH bands would severely limit the usability of the instrument in the red- and near-IR spectral regimes. Figure 8 shows the relative transmission of these three types of fibers for a 60-m length, as well as the theoretical upper limit for glass fibers.

Figure 8 Transmissive efficiency for three types of silica fiber optics with a length of 60 m.
End losses are not included. The dashed line indicates the maximum efficiency for glass fibers.

The photometric stability of fiber optics is one item of concern. The predominant application of the MOMFOS instrument will be to observe targets that are several magnitudes fainter than the nighttime sky. That implies that sky subtraction must be excellent to a part in a few thousand. Current general experience with fiber optic systems shows that great care must be exerted to achieve adequate sky subtraction quality at the level of only 1 part in 100. However, using an observing mode in which the fibers are switched between target and sky on a relatively short timescale of about 30 seconds per switch should allow excellent sky subtraction with fibers. This observing mode, nod-shuffle, in which the CCD detector shuffles the charge back and forth on the CCD in sync with the nodding of the telescope (or in this case, the tip-tilting of the second image relay mirror), is considered essential for MOMFOS observations. The only remaining concern is whether the fiber optic cables will be photometrically stable at the 1 part in a few thousand over the timescale of a few minutes. A static fiber cable should have no problem. However, the fibers in MOMFOS will be undergoing a gradual change as the telescope tracks and the field rotates. If there is any stiction of the fiber in its cable, the fiber may tend to undergo discontinuous rather than continuous relaxation as it changes shape. Such radical changes can modulate the propagating modes within the fiber. This in turn could show up in the spectrum as an additional noise term called modal noise. Baudrand et al.4 discuss modal noise limitations to the resultant signal-to-noise of an observed spectrum. In general, modal noise should only appear in high dispersion spectra with high signal to noise. The R = 18,000 mode of MOMFOS may be high enough so that such noise might become a limiting factor. Evaluation must be performed to verify that this effect does not significantly impact the photometric throughput of the fiber; otherwise, sky subtraction may be systematically limited.

Fiber Cable

The fiber optics will require adequate cabling to protect them from damage, the environment, and stress. Each fiber should be individually sheathed in a protective tube of Teflon or similar material. Thermal expansion joints will be required in the Teflon tubing at regular intervals to prevent stressing the fibers. Such thermal expansion joints are being implemented on the Hectospec fiber cable for the Multiple Mirror Telescope (MMT).

The fiber cable must interface with the prime focus fiber positioner in such a way that the fibers do not introduce any torque on the fiber probes, and the fiber probes do not stress the individual fibers. Likewise, the cable must interface with the spectrographs in a similar manner so as to keep fiber stresses at a minimum without compromising the optical alignment of the fibers with the instrument.

It is highly desirable to keep the fiber run continuous and to avoid any interfiber connections. Although there is some experience with interfiber connections in some fiber-fed instruments, it is our belief that the implementation of these interconnections causes more risk to the performance of the fiber than it reduces for maintenance issues. If it is deemed a requirement to have such an interfiber connector, it will be critical to evaluate the performance of prototype designs in order to minimize efficiency loss and stress introduced into the fibers.

Field Rotator

Because the telescope will not be on an equatorial mounting, the target field will rotate as the telescope tracks across the sky. The fiber positioner must therefore track in rotation as well. The most straightforward way to do this is to rotate the fiber positioner and WFS assemblies. The prime focus corrector optics would not rotate. The ADC assembly, however, must rotate in synchronization with the fiber positioner and sensors. Spectrograph

Figure 9 shows the schematic view of the spectrograph design fed by the fiber optic cables. Three such spectrographs are required to accommodate the 700 fibers.

Figure 9 Schematic view of the spectrograph design. The fibers are at the prime focus of the collimator
(on the left-hand side of the drawing). The light illuminates the grating (center). The camera images the
light on to a 4K by 4K CCD detector. The R = 1000 mode is shown (see Figures 10 and 11 for moderate and
high-resolution configurations).

Spectral resolving power and entrance aperture drive the ultimate beam diameter of the spectrograph. We selected a diffraction angle of 63 degrees for the R = 18,000 mode. An entrance aperture of 500 microns then sets the monochromatic beam diameter at 500 mm.

It is assumed that the CCD detector pixel size will be 15 microns. To get 2.5 pixel sampling of the 500-micron fiber would require a demagnification of 500/37.5 or 13.3, or a camera focal ratio of about f/0.4! Such cameras are not likely viable, so we either need to image slice the input image or oversample the resolution profile. For the current study, oversampling was selected over image slicing. Image slicing would require additional complexity in the optics at the fiber input, and would spread the light in the spatial direction by a factor of the number of slices. This would then require either a larger detector format and larger FOV in the camera or a larger number of spectrographs. Oversampling the resolution profile results in (1) a loss of spectral coverage, or (2) the need to increase the detector format in the spectral dimension and the FOV for the camera.

The camera was selected to have a target focal ratio of about f/0.75. This would demagnify the fiber size on to about 5 pixels. The detector could be binned by 2 pixels in the spectral direction in order to achieve optimal 2.5 pixel sampling. The focal ratio of the camera ended up at f/0.86 for the final concept design.

Fiber Interface

Of the 700 fibers, 233 will feed into each spectrograph at the prime focus location of a spherical collimator. The fibers will be divided into two groups, with the top half feeding in from above the instrument and the bottom half feeding in from below. The feed-in must not allow the fibers to introduce significant obstruction of the beam. The current design assumes an obstruction of 10 mm in width running across the full beam diameter. The fiber slit is 296 mm long, with a center-to- center spacing of about 1.27 mm between each fiber. This is to allow a blank gap to exist between the spectra on the detector so that nod-shuffle observing could be made without contamination between the spectra.

The fibers are mounted in parallel slots, with each fiber terminating along a spherical surface. This is to allow the fibers to lie along the curved focal surface of the collimator. Such an arrangement will require that each fiber be polished independently, or in small subsets rather than as a total unit.

Field Lens

A field lens is required to relay the pupil image to either the grating or to the camera. The current design has the pupil located between the first two lens elements of the camera assembly. Alternatively, the fibers could be individually tilted to relocate the spectrograph pupil to the desired location. Use of the field lens allows the fibers to be placed in parallel slots, eliminating the need to individually tilt each fiber.

Order Separation Filter

A slot can be located after the field lens for insertion of order separation filters. It may also be possible to put blocking filters within the grating assemblies as well.

Collimating Mirror

The collimating mirror is a spherical mirror with dimensions of 1020 by 500 mm. The long dimension is aligned parallel to the length of the fiber slit. The focal ratio of the mirror is selected to be f/4.5 rather than f/4.8. This allows compensation for residual focal ratio degradation in the fibers and also for the tilting of the fibers in the fiber positioner.

Articulation Arm

Volume phase holographic grating technology is the most desirable grating type for this instrument. Such technology requires that the spectrograph camera-to-collimator angle be adjustable in order to match the Bragg angles of the grating in use. This allows the transmissive gratings to always function at or very near to the Littrow condition. Although there are designs that use mirrors to do the articulation, the bench mounted aspect of the MOMFOS spectrographs make it relatively straightforward to actually articulate either the camera or the collimator. We decided to articulate the fiber input and collimator due to the smaller mass contained within that arm of the spectrograph, compared to articulating the camera and detector. The low dispersion mode is displayed in Figure 9. Figures 10 and 11 show the moderate and high dispersion configurations for the spectrograph.

Figure 10 Moderate dispersion (R = 5,000) mode for the spectrograph. The fibers and collimator
have been moved to an angle of 28.6 degrees to the grating normal. Figure 11 High dispersion (R = 18,000) configuration of the spectrograph with the collimator moved
to an angle of 63 degrees with respect to the grating.

Grating and Grating Mount

Volume phase holographic gratings were selected for their high efficiency and transmissive nature. This allows the camera and collimator to be placed as close as possible to the grating in a Littrow or close-to-Littrow configuration. The resultant reduction in anamorphic magnification helps to keep the size and complexity of the camera optics to a minimum.

The gratings range in size from 650 by 600 mm to as large as 650 by 1200 mm. This will require some effort to find a vendor willing to fabricate such large gratings. Table 1 lists the canonical set of 10 gratings needed for the instrument.

Table 1 Canonical set of 10 gratings needed for the instrument.
4336.2500.8337.7-662.61000650 by 600
2556.2850.3573.3-1125.21000650 by 600
191528.6500.1467.0-531.95000650 by 650
112628.6850.5794.2-904.65000650 by 650
445563.0400.0392.3-406.718000650 by 1200
356463.0500.0490.4-508.318000650 by 1200
287463.0620.0608.0-630.418000650 by 1200
240863.0740.0725.7-752.418000650 by 1200
209663.0850.2833.8-864.418000650 by 1200
187663.0950.0931.7-965.918000650 by 1200

The gratings will adequately function over a wide range of grating angles, allowing access of spectral regions not specifically indicated in the table, but giving total spectral coverage at all three resolution regimes over the optical band. As the gratings are tilted, the resolving power will undergo a slight change, especially for the high-resolution applications.

The gratings would be interchangeable on a rotary table that has its rotary axis concentric with the collimator articulation axis.

Camera Corrector Lenses

Three lenses make up the corrector lens assembly for the camera. The surfaces of all three elements are spherical. The first element is 740 mm in diameter, the second element is comparably sized at 850 mm in diameter, and the third element is 1000 mm in diameter. Glass types are UBK7 for the first and third elements, and the second element is made from fused silica. Although these elements are large, they are not expected to be a source of technical difficulty. It is expected that the AR coatings will be a MgF2/SolGel combination, to give low loss and broad-band performance.

Camera Mirror

The mirror in the camera is a spherical mirror with a radius of curvature of 1.256 m and a full diameter of 1.3 m. The coating on the mirror would be either a silver coating or an overcoated silver/aluminum coating such as that being developed by the Lawrence Livermore National Laboratories.

Field Flattener Lens

A doublet is currently serving in the design as a field flattener lens. Glass materials are UBK7 and LLF6. The back of the lens (nearest the CCD array) is plano. The lens has a full diameter of 150 mm.

Optical Performance of Spectrograph

Spot diagrams for the six modeled representative configurations of the spectrograph are shown in Figures 12-17.

Figure 12 Spot diagrams for R = 1000 mode from 338 to 663 nm. Figure 13 Spot diagrams for R = 5000 mode from 467 to 531 nm.

Figure 14 Spot diagrams for R = 18,000 mode from 490 to 508 nm. Figure 15 Spot diagrams for R = 1000 mode from 573 to 1125 nm.

Figure 16 Spot diagrams for R = 5000 mode from 794 to 904 nm. Figure 17 Spot diagrams for R = 18000 mode from 834 to 864 nm.

Figure 18 Quantum efficiency of the LBNL CCD detectors.

CCD Array

A 4K by 4K CCD array with 15-micron pixels serves as the detector for this instrument. For purposes of determining efficiency, the detector type was selected to be a fully depleted p-type CCD, as that developed by the Lawrence Berkeley National Laboratories. These CCDs have particularly excellent quantum efficiency redward of 600 nm, and up to 900 nm. They also do not suffer from fringing in the red as other types of CCDs do. Figure 18 shows the typical efficiency of this type of CCD.


There are two options for the Cryostat. The first would contain a rather large evacuated volume defined by the spherical mirror and the third camera corrector lens (the large bi-convex element). This would allow the central obstruction produced by the CCD and field flattener lens to be minimized. It would not require the field flattener to serve as a window for the CCD. Drawbacks include the large volume that must be evacuated and cooled, and the cemented doublet flattener lens might also be problematic for operation at -100 degrees C.

The second option is to utilize a small cryostat surrounding just the CCD array, and to use the field flattener lens as the cryostat window. A cold finger would connect this small chamber to the refrigerant system located to the side of the camera. An advantage of this approach is a smaller volume of space to be evacuated and cooled. A disadvantage is the possibility of a larger central obstruction due to the cryostat chamber. Such a cryostat might be similar to that being fabricated for the Gemini bHROS instrument. Calibration Unit

All instruments need a way to be calibrated, particularly spectrographs. A flat field and wavelength calibration are both required.

Projection Screen

Due to the nature of the prime focus corrector image relay, it will be possible to move a diffuser/screen near the location of the initial prime focus image as shown in Figure 2. The image of that screen will be relayed onto the fibers in the same manner that the image of the target field is relayed. The drawback to this approach is the requirement of excellent uniformity of illumination across the field, especially for flat field purposes. The ideal location for such a screen is a pupil image location.

Wavelength Calibration Lamps

A set of Helium, Neon, and Argon lamps could be utilized in the projection system, as well as a series of hollow cathode tubes containing Thorium-Argon, Copper-Argon, or any other preferred gasses.

Flat Field Lamps

Quartz lamps will also be housed within the projection system for flat field calibrations. OPERATIONAL CONCEPT

The MOMFOS instrument would be used in the following manner:

  1. The astronomers derive the astrometry for their targets prior to observing, and run the positions through a fiber assignment code that assigns the fiber optic probes into an optimal configuration that maximizes the number and/or importance of the targets to be observed for that configuration. This code will have to query for the time of the observations so that the effects of atmospheric refraction are appropriately taken into account.
  2. The spectrographs are configured for the desired spectral and resolution configuration by (1) inserting the appropriate grating, (2) rotating the grating to the optimal grating angle, and (3) articulating the fiber/collimator assembly to the appropriate angle. The cameras are focused, although this should be minimal because the spectrograph does not require any focus between different configurations, only for seasonal temperature compensation.
  3. The fibers are configured by first translating the fiber assembly back away from the focal surface. The acquisition TV system is inserted and used to monitor the fiber positions via a TV view system. Fiber configuration should take only about 5 to 10 minutes.
  4. The acquisition TV system is removed and the fibers are translated back onto the focal surface.
  5. The calibration system is moved into position for flat field and wavelength calibration. Note that it is critical that the fibers be configured prior to calibration, because the illumination of each fiber probe depends on that fiber's angular orientation to the optical axis. Recall that a fiber is positioned onto its target by a slight tip and/or tilt of the fiber. The maximum deviation of fiber axial alignment with the optical axis can be as large as two degrees. This will, in particular, have a potential effect on the flat field calibration. There may also be some second order effects on wavelength calibration.
  6. After calibrations, the calibration unit is removed from the optical axis.
  7. The fiber assembly is translated away and the acquisition TV assembly is inserted.
  8. The telescope is moved to the target field position, and an acquisition star is centered with the acquisition TV system.
  9. The ADC assembly is rotated to cancel out the atmospheric dispersion.
  10. The WFSs are inserted to acquire the requisite number of guide/wavefront stars.
  11. The telescope is focused and the AO system is initialized and started. It is assumed that the WFSs are also capable of maintaining telescope guiding on the target field.
  12. The acquisition TV is removed and the fiber assembly is reinserted.
  13. The observations are initiated.
  14. After a predetermined amount of time (on the order of 30 seconds), the nod-shuffle mode will shuffle the charge on the CCD and tip the tip-tilt mirror in the prime focus corrector to the off-target position. After another equal period of time, the CCD will shuffle the charge back and the tip-tilt mirror will return to the targets. These actions will continue until the end of the exposure.
  15. The data are read out and stored on a hard disk for quick look analysis.

Figure 19 Predicted system efficiency, excluding seeing losses, for the MOMFOS instrument in six
different configurations. PERFORMANCE ESTIMATE

Figure 19 shows the predicted efficiency for the instrument, excluding seeing losses on the fiber and atmospheric transmission loss. Peak efficiency in the red should approach 40%. These data assume a constant efficiency vs. wavelength for the air-glass surface losses. In reality, there will be a wavelength dependence of that loss, and the system could either be optimized for a specific wavelength with all air-glass interfaces peaking at the same wavelength, or averaged with each air-glass interface peaking at a different wavelength. Given the broad-band use of this instrument, it is more likely that the air-glass anti-reflective coatings will each be tuned for a different wavelength.

Seeing losses on the circular apertures of the fibers will vary as a function of the seeing. For 0.5" seeing, that loss will be about 45% for the 0.72" fibers. COST ESTIMATE

A cost estimate has not yet been made for this instrument. Uncertain cost drivers will be the prime focus corrector assembly, the adaptive mirror, the fiber positioner and fiber cable, and the large format VPH gratings.

Some items with associated estimated costs are: TECHNICAL ISSUES AND RISKS

The following items are identified as posing some level of technical risk.

  1. Gratings
  2. Fiber performance for sky subtraction
  3. AO correction for wind shake
  4. Detector cryostat packaging
  5. Fiber positioner
  6. Fiber relay optics Gratings

The large format volume-phase holographic (VPH) gratings envisioned for this instrument will require a significant amount of effort to find a source for their fabrication. The technical issues related to their fabrication are minimal, as holographic technology already exists for the fabrication of holographic elements up to 1 m in size. The challenge will mainly involve getting a vendor interested in implementing the required facilities to make the gratings, not developing the technology.

The Lawrence Livermore National Laboratory currently has 1-m holographic capability, but lacks the expertise for making holograms in dichromated gelatin (DCG). Their current capabilities are in photoresist, which is used for making holographically generated surface-relief gratings, not volume-phase holographic gratings.

Kaiser Optical Systems, Inc. (KOSI) in Ann Arbor, Michigan, is able to make and process DCG holograms as large as 380 mm on a side. Its current capability is limited to the production of head-up display units, not diffraction gratings. The facility for making diffraction gratings is limited to a monolithic size of about 100 by 100 mm. KOSI is currently exploring the possibility of making mosaiced gratings with a series of 100 mm elements on the same substrate. This might allow KOSI to make gratings of larger size, with a likely maximum size of 380 mm due to the limitations of its DCG processing facility.

Ralcon Labs in Utah is currently capable of making gratings as large as about 300 mm. It is unknown what size the facility is limited to with respect to its DCG processing facilities.

Centre de Spatial de Liege is currently learning how to make DCG gratings and will explore the fabrication of large format gratings up to about 300 mm during the next year. This effort is the result of a consortium of observatories interested in obtaining large format VPH gratings. Members of the consortium include ESO, NOAO, University of Michigan, GoLeM, and the AAO. Pending a successful outcome of this effort, a company will be spun off to continue making large format gratings on a commercial basis.

Chris Clemens, an astronomer at the University of North Carolina, is currently learning how to make DCG gratings. His intention is to make the gratings he needs for his instruments. He is considering expanding his effort to make large format gratings pending the results of his current exploratory effort. Other companies may eventually have the capability to make moderately large VPH gratings.

NOAO has a pair of high quality BK7 elements that were originally purchased for use as holographic collimators for KOSI in an effort to get them to upgrade their facilities. The glass still awaits generation, as KOSI is not currently interested. These elements can be made available for any other possible venue that might lead to the fabrication of large format VPH gratings. Chris Clemens has expressed possible interest in them. The glass is of a size that allows a clear aperture of 300 mm.

The MOMFOS instrument requires gratings that are slightly over a factor of two larger than that currently envisioned for large grating fabrication. It is likely that the gratings will have to be made of mosaics generated from 200- or 300-mm-sized gratings.

If VPH grating technology is not available, the spectrographs will need to be redesigned for use with classical, surface-relief gratings. Thermo RGL (formerly Richardson Grating Lab) will likely be able to make mosaiced gratings. However, for 2-dimensional mosaicing, RGL evidently must develop a new technique, because its current mosaic process only allows for the production of mosaics in the dimension along the dispersion axis of the grating.

A drawback to utilizing classical gratings is the requirement of making the gratings reflective rather than transmissive. This results in the inability to fully utilize the gratings in a Littrow configuration. A quasi-Littrow arrangement might be possible, but it will still lead to some complication in the spectrograph design. A future study should evaluate a design that utilizes reflection gratings. Fiber Performance for Sky Subtraction

Fiber optics have been notorious for their inability to do much better than 1% sky subtraction.5 However, use of a nod-shuffle observing mode has yielded good sky subtraction results in an experiment conducted by the Anglo-Australian Observatory using the Spiral and 2dF instruments.6 A fiber-fed system on the GSMT will most likely require the use of a nod-shuffle mode of observation.

The impact of requiring a nod-shuffle mode of operation includes the following:

Fiber Optic Photometric Stability

The 60-m long, 500 micron fibers will need to have photometric stability of one part in several thousand over a time scale of about 60 seconds in order for the nod-shuffle observing mode to adequately achieve the sky subtraction necessary to observe target objects that are six to seven magnitudes fainter than the night sky. Modal noise in the fiber as it twists about due to motion of the telescope and rotation of the instrument may be a limiting factor. Studies should be carried out to verify the level of photometric precision that fibers can deliver, and to determine what level nod-shuffle sky subtraction would work. AO Correction for Wind Shake

It will be critical to implement an AO mirror for wind shake correction. This is envisioned to be the first mirror in the prime focus image relay and corrector. The mirror has a diameter of 2 m and will likely need to have about 900 actuators. This is an obvious technical challenge. Detector Cryostat Packaging

The cryostat for the CCD detector is a technical challenge for either implementation. Fiber Positioner

The Echidna style fiber positioner is another technical risk. It appears that the AAO group will succeed in making the technology work for the Subaru telescope. Implementing it on the GSMT will require a slightly smaller spine footprint to allow the fibers to be spaced at 6-mm rather than 7-mm spacing, but this is probably not too significant of a risk.

The spine positions must not be disturbed by (1) vibrations introduced by the AO mirror, (2) the tip-tilt mirror, or (3) the action of moving the fiber assembly into and out of the image area. This will likely required some level of R&D effort to determine the susceptibility of the fibers to such vibrations.

The fiber assembly must rotate in order to track the target field. Such rotation must not cause the fiber spines to move off their targets. Fiber Relay Optics

The micro-lenses required to relay the f/1 prime focus light into the fibers will require high precision fabrication and alignment. Similarly, the lenses on the output end of the fibers will require comparable attention. The spectrograph optical design does not currently include the output micro-lenses. RECOMMENDATION FOR FUTURE STUDIES

The following future studies are recommended:

  1. Determine photometric stability of 60-m long, 500-micron diameter, multi-mode fiber to level of one part in 5000.
  2. Finalize spectrograph design to include micro-lenses on the output end of the fibers.
  3. Evaluate Echidna technology being developed at the AAO.
  4. Evaluate performance of Gemini bHROS CCD cryostat.
  5. Model image performance of prime focus corrector with AO correction of wind-buffeting and boundary layer turbulence. Determine number and magnitude of guide stars required.
  6. Continue efforts to find a vendor willing and able to fabricate large format VPH gratings.
  7. Explore use of reflective gratings as an alternative in the case that VPH gratings are unavailable. This will involve development of an alternative spectrograph design for the use of reflective gratings. It will also require a development effort with Richardson Grating Labs to develop the process for making 2-dimensional grating mosaics. REFERENCES

  1. Parks, R. J.; Barden, S. C. "Optimal Slit Size Study for the NOAO Next Generation Optical Spectrograph". BAAS 31, 1503 (1999)
  2. Glazebrook, K.; Bland-Hawthorn, J. "Microslit Nod-Shuffle Spectroscopy: A Technique for Achieving Very High Densities of Spectra". PASP 113, 197 (2001)
  3. Barden, S.C. "Review of Fiber-Optic Properties for Astronomical Spectroscopy". ASP Conf Ser 152, 14 (1998)
  4. Baudrand, J., Guinouard, I., Jocou, L., and Casse, M. "Use and Development of Fiber Optics on the VLT". ASP Conf Ser 152, 32 (1998)
  5. Wyse, R. F. G.; Gilmore, G. "Sky subtraction with fibres". MNRAS 257, 1 (1992); Barden, S.; Elston, R.; Armandroff; Pryor, C. "Observational Performance of Fiber Optics - High Precision Sky Subtraction and Radial Velocities". ASP Conf Ser 37, 223 (1993)
  6. Cannon, R. "Optimal Sky Subtraction with Fibre Systems". AAO Newsletter 96, 13 (2001)

March 2002