KAOS: kilo-aperture optical spectrographSamuel C. Bardena*, Arjun Deyb, Brian Boylec, Karl Glazebrookd

aAnglo-Australian Observatory, 167 Vimiera Road, Eastwood NSW 2122, AustraliabNational Optical Astronomy Observatory, 950 N. Cherry Ave., Tucson, AZ 85719, USA

cAustralia Telescope National Facility, PO Box 76, Epping NSW 1710, AustraliadDept of Physics and Astronomy, Johns Hopkins University, 3400 N. Charles St., Baltimore, MD

21218, USA

ABSTRACT

A design is described for a potential new facility capable of taking detailed spectroscopy of millions of objects in theUniverse to explore the complexity of the Universe and to answer fundamental questions relating to the equation ofstate of dark energy and to how the Milky Way galaxy formed. The specific design described is envisioned forimplementation on the Gemini 8-meter telescopes. It utilizes a 1.5° field of view and samples that field with up to~5000 apertures. This Kilo-Aperture Optical Spectrograph (KAOS) is mounted at prime focus with a 4-elementcorrector, atmospheric dispersion compensator (ADC), and an Echidna-style fiber optic positioner. The ADC doublesas a wobble plate, allowing fast guiding that cancels out the wind buffeting of the telescope. The fibers, which can bereconfigured in less than 10 minutes, feed to an array of 12 spectrographs located in the pier of the telescope. Thespectrographs are capable of provided spectral resolving powers of a few thousand up to about 40,000.

Keywords: Astronomical instrumentation, spectrographs, telescopes, Gemini observatory, multi-object spectroscopy

1. INTRODUCTION

The US astronomical community held a workshop in October 20001 to discuss the optical/infrared ground-based systemof public and privately run US facilities. At that workshop, it was concluded that additional emphasis was needed forwide-field multi-object spectroscopy (MOS) on 8-10 meter class telescopes beyond what was already envisioned at thattime. The National Optical Astronomy Observatory (NOAO) followed up on that recommendation a year later with aworkshop specifically dedicated to the topic of wide-field MOS on large aperture telescopes with a confirmation of theneed for such facilities2. It was then decided that a concept for a very wide field of view should be developed with thegoal of this facility residing on the Gemini 8-meter telescopes.

Contemporaneously, Gemini initiated its own process to define the next generation suite of instrumentation (aka theAspen process). The science cases were made to justify development of a wide-field MOS facility for the Geminitelescopes. The final specifications of such a facility call for a field of view 1.5° in diameter with the ability to doeither high or low dispersion optical (0.4 to 1.0 micron) spectroscopy on up to 1000 or more simultaneous targetswithin that field. Gemini is currently initiating a feasibility study for this instrument with a goal of actual developmentduring the second half of the decade and first use early in the next decade.

The paper presented here discusses the instrumental concept, known as KAOS3, developed as a result of the NOAOworkshop4. The development of this concept helped produce the science cases that were presented in the Aspenprocess that ultimately led to the decision of Gemini to explore such a facility. The KAOS concept also serves as thestrawman design that Gemini is using to further examine the concept.

* Email scb@aaoepp.aao.gov.au; phone +61 2 9372 4852; fax +61 2 9372 4880; web www.aao.gov.au

2. KAOS

KAOS (Kilo-Aperture Optical Spectrograph) is an instrument that provides a large field of view on an 8-meter classtelescope with the ability to simultaneously observe the optical spectra of up to ~5000 targets at either high or lowresolution. The performance goals of the instrument are:• 1 to 2° field of view• High density of targets per exposure (>1000)• High efficiency (>30% peak)• Spectral window of 0.39 to 1.5 microns

• Optical channel of 0.39 to 1.1 microns• Non-thermal IR channel of 1.1 to 1.5 microns as a future upgrade

• 1 arc-second apertures• Spectral resolution range of 2000 to 30,000• Ability to do Nod&Shuffle5,6,7 observations

The Gemini telescopes have the following constraints:• f/1.7 primary• 1.5 meter space between uncorrected prime focus and the dome• wind shake amplitude of ~1 arc-second• wind shake rate of ~2 Hz• Spectrograph lab in the pier requiring ~60 meter fiber length

Gemini has limited space available for a prime focus instrument. Hence, slit-fed instruments are not likely to fit.

The windshake issue is of particular concern for a prime-focus mounted instrument. The fast tip-tilt f/16 secondary thatcurrently feeds the instruments mounted at the Cassegrain port automatically stabilizes the image from wind buffeting.A transmissive wobble plate is envisioned for use in the KAOS instrument similar to that implemented in theMEGACAM instrument at the CFHT8.

2.1 Science objectivesKAOS is a concept for a facility instrument with numerous scientific applications. Two particular applications drive thespecifications of the instrument:• A survey of the distribution of ~1,000,000 galaxies in the z=0.5-3 redshift regime to detect and measure the

underlying structure of the Universe created and driven by the force of the unknown dark energy that is apparentlyaccelerating the expansion of the Universe9,10,11. The effort to study and understand the nature of this energy isconsidered to be one of the most fundamental and perplexing problems in Physics this century12.

• A survey of ~150,000 stars in our own Milky Way galaxy at high dispersion in order to chemically tag them andmeasure their space velocities (in conjunction with proper motion and parallax surveys) so that we can trace theirorigins and piece together the processes that led to the formation of the galaxy that we see today13.

2.2 Wide field corrector, ADC, and wobble plateThe objective of the KAOS design is to exploit as wide a field as is reasonably possible given the various constraints ofsize and mass allowed at the top end of the Gemini telescopes. Although a 2° field of view corrector could be opticallydesigned14 with 0.5 to 1 arc-second image quality, the optics are large (1.5 meter diameter front element) and massive(~1700 kg of glass). It was decided to scale back to a 1.5° field of view.

Figure 1 shows a working optical design for the corrector with a functional ADC that also serves as a wobble plate forimage stabilization against wind buffeting. The corrector is a 4-element design with Fused Silica and BK7 elements. Inorder to better match the numerical aperture of the fibers, the corrector has some built in magnification to convert thef/1.7 focal ratio from the primary to about f/2.4. Most surfaces are spherical with the exception of the three aspheresindicated in the figure. The ADC consists of a pair of prism sets consisting of BK7 and LLF6 glass designed to operateup to 70° Zenith angle across the 0.39-1 micron spectral window. Both sets of prisms can also be tip-tilted to stabilizethe image against the wind and to nod the images in a Nod&Shuffle mode. Details of spot diagrams, image stabilizationperformance, and other possible design options are described further by Liang, Barden, and Robles14. In essence, the

corrector is capable of correcting atmospheric dispersion at a Zenith angle of 70° and can stabilize image motion with anamplitude of a few arc-seconds with 80% encircled energy contained within 0.6” over most of the field. If the ADC canbe implemented within the powered elements of the corrector rather than as a separate unit, then image quality could beas good as 80% encircled energy within 0.5” over the majority of the 1.5° field.

Figure 2: Solid model schematic of KAOS corrector assembly showingvarious components and the space envelope available for the instrumentpackage.

Figure 1: Schematic of one possible design for the KAOS wide field corrector, ADC, and wobbleplate. The aspheric surfaces are indicated by the *.

Figure 2 displays a solid model schematic of a corrector assembly that fits within the space constraints of the Geminitelescopes. The focal plane instrumentation must fit within a volume that can be no more than 0.88 meters in depth.This space limitation most likely eliminates the possibility of using slit-fed spectrographs, but a fiber positioner can beused to feed instrumentation located elsewhere.

2.3 The fiber positionerThe key technology that makes KAOS a realistic concept is the fiber positioning technology developed by the Anglo-Australian Observatory for the FMOS instrument on the Subaru telescope15,16,17. The fibers are mounted in tip-tiltablespines rather than on magnetic buttons with prisms as in most fiber positioners such as OzPoz18, 2dF19, Hectospec20, andHydra21. This novel implementation allows a very high density of fibers to be mounted within the focal image area. Thefibers each patrol a relatively small region on the sky and are moved within that region by the tip-tilt action of the spine.Figure 3 shows how the fibers are arranged and their range of motion. It should be noted that the tip-tilt action doesintroduce an angular deviation of the fiber with respect to the incoming beam of light. This limits the application of thisapproach to relatively fast focal ratios (<f/3) such as that encountered at prime focus. The uniform spacing of the fibersmakes this approach a good match for all-sky surveys of stars and galaxies, but is not necessarily ideal for surveyingdensely clustered systems such as globular clusters.

The nominal fiber spacing on the FMOS Echidna positioner is 7 mm. This spacing would allow about 5000 fibers to besituated within the 1.5° field of view generated by the wide field corrector discussed in the previous section. With aplate scale of about 11-12 arc-seconds per mm, each fiber is able to patrol a region on the sky 2.5-2.8 arc-minutes indiameter. Neighboring fibers could be placed to within at least 30 arc-seconds of each other and, depending on finalspine construction, might possibly be placed as close together as 5 to 10 arc-seconds.

Figure 3: Close up schematic view of fiber arrangement and range of motion.

One of the drawbacks for most fiber positioners is the largetime required to configure the fibers. Since the Echidnatechnology has individual actuators, the configuration timefor the 5000 fibers can be quite short depending on thecontrol and technique for positional feedback. It isanticipated that the KAOS fibers could be reconfigured onthe timescale of a few minutes.

2.4 Fiber cable and routingGemini delivers ~0.6 arc-second median seeing at either ofits two sites. This implies that 1 arc-second diameter (~100micron) fibers may be utilized depending on the performanceof the corrector. Roughly 5000 fibers can be implemented inthe field. The location of the suite of spectrographs in thepier lab implies a fiber cable length of about 60 meters.Figure 4 gives the typical transmission performance for avariety of fiber types. It should be noted that perfect fiberswould still lose light in the blue due to Rayleigh scattering.Given the anticipated fiber performance, the KAOS

instrument will generally be limited to the 0.36-1.7 micron wavelength band. The blue end being limited by losseswithin the fiber and the red end being limited by thermal background.

One of the challenges for implementation of KAOS willbe the cabling of the 300 km of fiber. An interfiberconnector will be required in order to allow relativelystraightforward exchange of the telescope top ends.Figure 5 shows a solid model rendering of the telescopewith KAOS installed. The fiber connector would belocated at the existing interface of the telescope maintruss and top end as indicated in the figure. Gemini wasdesigned to allow top end changes. It is imagined thatthe 5000 fibers from KAOS would be divided intogroups that follow three of the four main trusses (one ofthe four paths is dedicated for use by the telescope laserguide star system). The cables enter the telescopesupport structure near the elevation axis where they areencased inside the tube structure to a point underneaththe dome floor. From there, the cables snake to a holelocated centrally on the Azimuthal axis and drop downinto the pier. The cables are then distributed out to thespectrographs located inside the pier lab.

2.5 Spectrograph conceptsThe spectrographs for an instrument like KAOS could bedesigned in a wide variety of approaches. However, thecost of implementation and support will likely restrict therange of options. Ideally, the spectrographs might betransmissive designs that minimize light lost due to anycentral obstructions. Unfortunately, all transmissivespectrographs with wide fields of view, large detectors,and a wide range of wavelength operability tend to havenumerous optical components and run the risk ofrequiring exotic materials and/or aspheric surfaces thusrendering them as very expensive to build. Although aformal design study should include such spectrographs, it

Figure 5: Solid model showing the Gemini telescope with theKAOS top end and the spectrographs located in the pier. Thepathway for one of the fiber cables is shown.

Figure 4: Transmission efficiency for 3 different fiber typesover 60 meter lengths.

was decided in the KAOS study to explore what is likely a less expensive approach that still delivers adequate, thoughpossibly not optimal, performance. It was also presumed that a single, all performing spectrograph could not be builtwithin reasonable cost (though this should not be ruled out in a formal design exploration).

In order to allow Nod&Shuffle5,6,7 observing for the faint surveys, there must be adequate space in between the imagedspectra on the detector to shuffle the spectra, record the sky spectra, and achieve clean separation of the spectra. A 10-pixel spacing was assumed for fiber spectra imaged onto 3 spatial pixels. A typical 4k detector is therefor able to imageabout 400 fibers or 800 total Nod&Shuffled spectra. Thus, about 12 spectrographs would be required to image thespectra from all of the fibers.

The simplest spectrograph design is typically that with reflective collimators and cameras. The fiber slit can be made tohave a fairly minimal mechanical footprint in the collimator light path. Schmidt-like cameras can produce very widefields of view with fast focal ratios and acceptable central obstructions. However, it was decided that the costs might belowered further by keeping the surfaces of all of the optical components spherical, if the number of transmissivecorrecting elements could be kept at a reasonable number. This was identical to the approach taken in the study of asimilar concept for a multi-object spectrograph for the GSMT 30-meter telescope22,23 in which the spectrograph opticswere restricted to all spherical surfaces on 2 mirrors (collimator and camera), three correcting elements, a field lens onthe fibers, and a field flattening lens on the detector.

Figure 6 displays a schematic and solid model view for a spectrograph concept that effectively meets the majority ofrequirements for KAOS. It is a 200 mm beamed instrument with spherical mirrors in the collimator and the camera. A5-element corrector has all spherical surfaces and is achromatic across the 0.39-1.0 micron spectral window givingbetter than 2 pixel 50% encircled energy. The input fibers are 150 micron1 arc-second and illuminate the f/4 collimator.It is assumed that the fiber connector on the telescope relays the f/2.4 input focal ratio to f/4 with appropriate relayoptics within the connector. Volume-phase holographic (VPH) gratings are assumed with line frequencies ranging from450 to 3500 l/mm to give spectral resolving powers (l/Dl) of 2000 to 20,000. Note that the fiber/collimator arm isassumed to articulate rather than the heavy camera.

Although the efficiency performance of this design is quite good as shown in Figure 7, the obstruction by the detectorpackage is significant and the resolution doesn't quite achieve the maximum level desired (l/Dl of 30,000).Improvements can likely be achieved by altering the approach by any of the following:• Increase the beam diameter to 300 mm so that the central obstruction becomes smaller and so that the resolving

power is also increased by a factor of 1.5 for a given grating line frequency.• Explore transmissive or off-axis designs for the collimator so that the fiber interface can be simplified and not

obstruct any of the light.

Figure 6: Schematic and solid model for a possible KAOS spectrograph. The collimator is f/4. All surfaces are spherical. Thedetector is a 4k by 4k format CCD. The fibers and collimator articulate about the VPH gratings.

• Explore transmissive or off-axis camera designs to minimize and/or eliminate the central obstruction loss of thedetector package.

• Explore the use of traditional Echelle gratings and use an R4 Echelle to get l/Dl of 30,000.

Increasing the beam diameter is probably the easiest and most beneficial change in the design. Not only does theefficiency increase due to the smaller footprint, but a larger detector (4k by 6k) can be used to allow 600 fibers perspectrograph. The total number of spectrographs could be reduced from 12 to 8.

It was also considered in the development of the KAOS concept that it may be desirable to assign and dedicate groups offibers to different types of spectrographs. The simplest division would be to have some number of the fibers feed intohigh-dispersion spectrographs and the rest feed low dispersion spectrographs. The faint object science could probablybe done with 2/3rd the total number of available fibers. Sky coverage is still complete with a factor of two redundancy(ie. each target could be reached by two fibers). The remaining 1/3rd fibers would feed into a bank of high dispersionspectrographs for observation of brighter objects. The density of targets in the field would be sufficiently matched eventhough only 1 fiber could reach any given target. This division of fibers would allow two different spectrograph designsthat might allow better optimization of performance and savings in cost.

The low dispersion spectrographs could be smaller beamed instruments, perhaps 50 to 100 mm beam diameter, allowingthem to be produced with all-transmissive optics giving better efficiency due to the removal of central obstructions in thebeam. Alternatively, the number of larger beamed, high dispersion spectrographs could be reduced to 3 or 4 giving thepossibility of further cost reductions by minimizing the number of large optics and high dispersion gratings.

Figure 7: Total estimated system efficiency for KAOS. This includes the telescope, corrector,fibers, spectrograph, gratings, and detector. It does not include seeing losses. Six gratingconfigurations are displayed. The thick, solid line is the efficiency without the gratings or filters.

The currently envisioned location for the spectrographsis in the pier lab as indicated in Figure 5. Thespectrographs will be bench-mounted and must bereadily accessible by support staff for maintenance andconfigurations. Figure 8 displays a close up view ofhow the spectrographs might be installed. Here thereare 12 spectrographs (only 9 shown in the figure)arranged in four stacks or bunks.

3. CONCLUSIONS AND CURRENT STATUS

The concept described here suggests that it istechnically feasible to build a 1.5° field of view, ~5000target, multi-object spectrograph for the Geminitelescopes. The wide field corrector appears to give thelargest risk with the large and massive opticalcomponents. However, the fabrication of such elementsis currently within the realm of that being considered for

other astronomical facilities such as the Large Synoptic Survey Telescope (LSST)24.

Gemini has concluded its Aspen process and is currently initiating a feasibility study for a Gemini Wide Field, Fiber-fed,Optical, Multi-Object Spectrograph (WFMOS) that has specifications essentially identical to those of the KAOSconcept. It is anticipated that the studies will lead into a conceptual design phase during 2005 and that such a facilitywill be in science operation by the early part of the next decade.

ACKNOWLEDGMENTS

The KAOS concept and scientific application was developed by a large group of collaborators. Further information canbe obtained from http://www.noao.edu/kaos. Members of the original KAOS team are: Sam Barden (NOAO), ChrisBlake (UNSW), Joss-Bland Hawthorn (AAO), Brian Boyle (ATNF), Michael Brown (NOAO), Matthew Colless(AAO), Warrick Couch (UNSW), Len Cowie (IfA/UH), Arjun Dey (NOAO), Daniel Eisenstein (Univ. of Arizona), KenFreeman (MSO/ANU), Karl Glazebrook (Johns Hopkins Univ.), Charles Harmer (NOAO), Buell Jannuzi (NOAO),Francis Keenan (Queens University, Belfast), Rolf Kudritzki (IfA/UH), Ming Liang (NOAO), Eric Linder (LawrenceBerkeley Laboratory), Anna Moore (formerly of the AAO), David Meyer (Northwestern Univ.), Jeremy Mould(NOAO), Joan Najita (NOAO), Robert Nichol (Carnegie Mellon Univ.), Knut Olsen (NOAO), John Peacock (ROE),James Robinson (NOAO), Rick Robles (NOAO), Hee-Jong Seo (Univ. of Arizona), Thaisa Storchi-Bergmann (UFRGS,Brasil), Steve Strom (NOAO), Nick Suntzeff (NOAO), Alex Szalay (Johns Hopkins Univ.), Rosie Wyse (Johns HopkinsUniv.).

REFERENCES

1. Alcock, C., Beichmann, C., Boroson, T., Crocker, J., Dressler, A., Gunn, J., Illingworth, G., Kirshner, R., Lunine, J.,McKee, C., Mushotzky, R., Osmer, P., Peoples, J., “Report of the First Workshop on The Ground-Based O/IRSystem”, http://www.noao.edu/gateway/oir_workshop/report.pdf, 2000.

2. Brown, M.J.I., and Dey, A., “Next Generation Wide-Field Multi-Object Spectroscopy”, Astron. Soc. PacificConference Series, 280, 2002.

3. Dey, A., and Boyle, B., “KAOS Purple Book”, http://www.noao.edu/kaos/KAOS_Final.pdf, 2003.4. Barden, S.C., “A Wide Field Concept for an 8-meter Telescope”, Astron. Soc. Pacific Conference Series, 280, pp

89-96, 2002.5. Cuillandre, J.C., Fort, B., Picat, J.P., Soucail, G., “Back and forth spectroscopy: optimization of an optical nod-and-

shuffle technique to reach fainter objects and increase the multiplex gain on multi object spectrographs”, Proc. SPIE,4841, pp. 1531-1538, 2003.

Figure 8: Close up view of three of the four spectrograph bunkslocated in the pier lab.

6. Glazebrook, K., and Bland-Hawthorn, J., “Microslit Nod-Shuffle Spectroscopy: A Technique for Achieving VeryHigh Densities of Spectra”, Pub. Of Astron. Soc. Pacific, 113, pp. 197-214, 2001.

7. Abraham, R.G., et al., “The Gemini Deep Deep Survey. I. Introduction to the Survey, Catalogs, and CompositeSpectra”, Astron. Journ., 127, pp. 2455-2483, 2004.

8. Boulade, O., Vigroux, L., Charlot, X., Borgeaud, P., Carton, P.H., de Kat, J., Rousse, J.Y., Mellier, Y., Gigan, P.,Crampton, D., and Morbey, C., “MEGACAM, the next generation wide-field camera for CFHT”, Proc. SPIE, 3355,pp. 614-625, 1998.

9. Riess, A.G., et al., “Observational Evidence from Supernovae for an Accelerating Universe and a CosmologicalConstant”, Astron. Journ., 116, pp. 1009-1038, 1998.

10. Blake, C., and Glazebrook, K., “Probing Dark Energy Using Baryonic Oscillations in the Galaxy Power Spectrum asa Cosmological Ruler”, Astrophys. Journ., 594, pp. 665-673, 2003.

11. Seo, H.-J., and Eisenstein, D.J., “Probing Dark Energy with Baryonic Acoustic Oscillations from Future LargeGalaxy Redshift Surveys”, Astrophys. Journ., 598, pp. 720-740, 2003.

12. Turner, M.S., et al., Connecting Quarks with the Cosmos: Eleven Science Questions for the New Century, TheNational Academies Press, Washington DC, 2003.

13. Freeman, K., and Bland-Hawthorn, J., “The New Galaxy: Signatures of Its Formation”, Ann. Rev. Astron. andAstrophys., 40, pp 487-537, 2002.

14. Liang, M., Barden, S.C., and Robles, R., “Wide field corrector with ADC compensator and image stabilizer for f/1.7Gemini telescope prime focus”, Proc. SPIE, 5492, these proceedings, 2004.

15. Gillingham, P., Miziarski, S., Akiyama, M., and Klocke, V., “Echidna – a Multi-Fiber Positioner for the SubaruPrime Focus”, Proc. SPIE, 4008, pp. 1395-1403, 2000.

16. Gillingham, P., Moore, A., Akiyama, M., Brzeski, J., Correll, D., Dawson, J., Farrell, T., Frost, G., Griesbach, J.,Haynes, R., Jones, D., Miziarski, S., Muller, R., Smedley, S., Smith, G., Waller, L., Noakes, K, and Arridge, C.,“The Fiber Multi-Object Spectrograph (FMOS) Project: the Anglo-Australian Observatory role”, Proc., SPIE, 4841,pp. 985-996, 2003.

17. Brzeski, J.K., Gillingham, P.R., Moore, A.M., Smith, G., Muller, R., Smedley, S., Correll, D., and Dawson, J,“Echidna: the engineering challenges”, Proc. SPIE, 5492, these proceedings, 2004.

18. Gillingham, P.R., Popovic, D., Farrell, T.J., and Waller, L.G., “The performance of OzPoz, a multi-fiber positioneron the VLT”, Proc. SPIE, 4841, pp. 1170-1179, 2003.

19. Taylor, K., Cannon, R.D., and Watson, F.G., “Anglo-Australian Telescope's 2dF Facility”, Proc. SPIE, 2871, pp.145-149, 1997.

20. Fabricant, D.G., Hertz, E.N., Szentgyorgyi, A.H., Fata, R.G., Roll, J.B., and Zajac, J.M., “Construction of theHectospec: 300 optical fiber-fed spectrograph for the converted MMT”, Proc. SPIE, 3355, pp. 285-296, 1998.

21. Barden, S.C. andArmandroff, T., “Performance of the WIYN fiber-fed MOS system: Hydra”, Proc. SPIE, 2476, pp.56-67, 1995.

22. Barden, S.C., “Wide-field optical spectrograph for a 30-meter telescope”, Proc. SPIE, 4841, pp. 1760-1770, 2003.23. Strom, S. et al., “Enabling a Giant Segmented Mirror Telescope for the Astronomical Community”, http://www.aura-

nio.noao.edu/book/index.html, 2002.24. Seppala, L.G., “Improved optical design for the Large Synoptic Survey Telescope (LSST)”, Proc. SPIE, 4836, pp.

111-118, 2002.