Annu. Rev. Astron. Astrophys. 1991. 29: 499-541 Copyright © 1991 by Annual Reviews. All rights reserved

4.1 Multiple-Object Optical Spectroscopy

Surveys demand more efficient use of the corrected field of view of a telescope than traditional single-slit spectroscopy needs. As surveys shift towards the study of fainter galaxies with surface number densities larger than a few per square degree, the multiplexing advantage of multiple-object spectrographs (MOSs) encourages long integrations that would be uneconomical for a single object. MOSs have developed rapidly in the last decade, and hold the promise for an order of magnitude increase in the growth of the extragalactic radial velocity data base in the near future. We discuss separately advantages and limitations of these techniques, including a brief overview of the first effective MOS, the objective-prism Schmidt. Ellis & Parry (1988) have written an excellent comparative review of multiple-object spectroscopy.

OBJECTIVE-PRISM SPECTROSCOPY Markarian and coworkers (Markarian 1967 and Markarian et al 1981 for other references) conducted a highly successful objective-prism survey using the 102/132-cm Byurakan Schmidt telescope, producing a list of about 1500 objects with ultraviolet continua. This effort has been followed most notably by the Cerro Tololo objective-prism survey carried out with the 61/91-cm Schmidt at CTIO (Smith 1975, MacAlpine & Williams 1981 and references therein). The objective-prism is a quick survey technique, which provides large-scale sky coverage of relatively bright sources (B ~ 17 for the Byurakan survey, B ~ 18.5 for the Tololo survey) with strong emission lines, whereby multiplexing is obtained thanks to the large field of the Schmidt. The near-UV to blue spectra are of very low dispersion (2500 Å mm^-1 and 1740 Å mm^-1 at H, respectively, for the Byurakan and Tololo surveys), and in some of the sources they can yield a redshift determination with an error of about 0.03 in z. This accuracy is only of moderate interest, at best, in large-scale structure studies. However, Cooke et al (1981) underscore that measurements on UK Schmidt plates of the 4000-Å break can provide radial velocities that are accurate to ~ 1800 km s^-1 for galaxies with b_J < 18.7. Parker et al (1987) propose that this feature can be profitably used in the study of the large-scale distribution properties of early-type galaxies; Beard et al (1986) have adopted a similar technique to obtain estimates of 496 redshifts - accurate to 2000 km s^-1 - to m_J ~ 17.8, in a UKST field in the Indus supercluster region.

OPTICAL FIBER SPECTROGRAPHS Optical fiber spectrographs typically can convey light simultaneously from 10 to 100 sources over field sizes of typically 0.4°-1° in diameter. Thus, a fiber MOS is best suited for the study of sources with surface number densities of 100 deg^-2 or higher. In earlier systems, the coupling of fibers and source locations in the focal plane was slow and cumbersome. However, the development of automated fiber-positioning devices allows increasingly fast repositioning of the fiber assembly in the focal plane, and thus makes modern multiple fiber spectrographs practical instruments in fields of lower source surface density. The tendency to build devices with increasingly larger numbers of fibers is complemented with that of obtaining systems that can be repositioned readily for different fields in the course of an observing session. Hill (1988) chronicles a highly readable history of the first ten years of multiobject fiber spectroscopy.

The first fiber MOS was developed at the Steward Observatory. Several other systems were built at major observatories in the early 1980s, most notably FOCAP (Gray 1984) for the AAT, OPTOPUS (Lund & Enard 1984) at ESO, and Nessie at Kitt Peak (Barden & Massey 1988). These systems rely on the preparation of aperture plates, with holes drilled with high precision at the locations of the target sources. Prior to each observation, the fiber ends must be attached to the plate. The procedure has important limitations, for it demands good quality astrometric information, and therefore considerable advance preparation: The observing programs are not easily adapted to schedule changes made necessary by prevailing conditions, aperture plate changes are very slow, and handling of fibers may produce considerable wear and tear on the fiber ends. As for all MOSs, the exposure times are those necessary for the weakest source in the field. In the late 1980s new automatic fiber positioning devices were built, such as the MX system at Steward (Hill 1988) and Autofib at the AAT (Parry & Sharples 1988). These two systems illustrate two different mechanical approaches to the problem. MX can position 32 fibers over a 45' field, using 32 independent robotized arms that intrude in the field from its periphery (like fishing poles in a pond). In the case of Autofib, a single very fast robot sequentially positions 64 fibers over the 40' Cassegrain field of the AAT. The fibers enter the field parallel to the focal plane and are kept magnetically in position. Robotized systems have been built or are about to enter operation for most major telescopes. Ingerson (1988) discusses design trade-offs applicable to fiber MOSs, as they led to the decisions affecting the construction of the Argus system at the 4-m telescope at CTIO.

Positioning speed is maximized in the MX design: A given configuration can be attainable in slightly over a minute. Positioning via a single robot, as in designs of the Autofib type, typically take several minutes. On the other hand, many independent robotized arms get in each other's way, so that the number of fibers can be higher in Autofib-type systems: MX, Argus, and Decaspec at the 2.4-m Hiltner telescope (Fabricant & Hertz 1990) can aim respectively at 32, 24, and 10 sources, while Autofib and the Norris spectrograph for the 5-m Palomar telescope (Cohen et al 1988) can do so at 64 and 100 sources, respectively. The physical size of the arms in one case, and that of the magnetic buttons used to keep fibers in position in the other, limits the minimum distance at which fibers can be set. In both cases, designs have been obtained where such separation has been reduced to 15" or less. In some designs, positions of fibers can be seen on the acquisition/guide TV camera and, with multiple-arm systems, interactive autocentering is possible after the source configuration has been roughly acquired. The increasingly good transmission quality of fibers makes it possible to locate spectrographs and detectors in a remote, immobile environment, rather than on the telescope, thereby improving instrument stability and reducing flexure worries. At this time, fiber losses and those related to focal ratio degradation make an observation through an optical fiber slower by approximately a factor of 2 than those made with a long slit (Fabricant & Hertz 1990).

Fibers used in MOSs have wavelength coverage limitations: dry (i.e. pure) fibers transmit well in the red and infrared, not in the blue; wet fibers (which are doped with small amounts of OH^-1) transmit well in the blue, but not in the red. Thus, a choice of the usable spectral region is set by the fibers installed in the MOS. Typically, the spectrograph detector is a CCD; spectra from each fiber need to be separated by at least 6-8 pixels from each other, perpendicular to the dispersion direction, in order to prevent contamination of light between fibers. The size of the CCD thus sets an upper limit to the numbers of fibers in a MOS. Fiber MOS sky subtraction characteristics are restrictive, especially with very faint objects, unless a large fraction of the fibers is devoted to acquiring sky photons.

Fiber MOSs have also been implemented for use with Schmidt telescopes, making the fiber technique attractive for objects with surface-number densities of less than a few per square degree. The notable example is the FLAIR system at the UK Schmidt at Siding Spring (Watson et al 1988), with 35 fibers. Parker & Watson (1990) have reported successful performance of FLAIR with 116 redshifts to a magnitude limit of m_J = 16.8. Current limitations in performance will be greatly improved with the forthcoming 100-fiber system FLAIR-2.

MULTISLIT SPECTROGRAPHS An effective use of the corrected field can als be obtained by placing several slitlets at the positions of desired sources in the focal plane, coupled with an imaging camera with a dispersing element such as a transmission prism in its collimated beam. This technique is restricted to smaller fields of view - those of the imaging camera, which is usually a CCD chip - and provides lower dispersion than that obtained with fiber MOSs. The dispersed sky photons overlapping the object spectrum, as in the case of objective-prism plates, can be avoided by having the slits placed in an aperture plate.

As in the case with fiber MOSs, technical developments toward having automatic positioning of slitlets in the field of view have been pursued. However, this path has not led to a general-purpose instrument configuration yet, as discussed by Ellis & Parry (1988), and the pre-etched aperture plate method is preferred.

Because sky subtraction is generally more effective with slits than with fibers, slit MOSs can target fainter sources, and they can thus simultaneously obtain a fair number of targets in spite of their relatively small fields of view. The EFOSC at the ESO and the Low Dispersion Survey Spectrograph (LDSS) commissioned in 1988 at the ATT represent two successful versions of the slit MOS. The LDSS is endowed with a relatively large field (12'), which, when coupled with a large CCD camera, such as the Tek 2048², should allow simultaneous placing of about 100 spectra, as opposed to 25-30 spectra with 1980s variety CCD chips. Colless et al (1990) describe both the instrument and the first results of a deep LDSS redshift survey. Many characteristics of multislit MOSs are better illustrated vis-a-vis those of fiber MOSs.

Several variables determine the relative quality of the performance of fiber and slit MOSs. Among them, the multiplex gain - or the number of spectra that can be acquired - can be higher for fiber than for slit MOSs. The quality of the sky subtraction, on the other hand, tends to be superior in slit MOSs because sky background is acquired in the immediate vicinity of the object (however, some fiber systems like the Norris at Hale and Decaspec at Hiltner, have satisfactory solutions to this problem). The size of the field of view can be much larger in fiber MOSs, which can sample the whole field of view of the telescope rather than just that physically spanned by the detector. Additional concerns involve sensitivity, ease and the agility of operation, and the need for elaborate preparation and versatility.

The inroads made in robotized systems with many dozen fibers and the relaxation of the need for high-quality astrometric data prior to the observations (allowing interactive adjustments to fiber positioning) have been important in making fiber MOSs the systems of promise for the next decade. The lack of agility in the observations and the need for elaborate and sometimes uneconomical preparation required by systems with specially manufactured aperture plates impose the strongest inhibition to their extensive use. To interested outsiders in this field, such as the present writers, it appears that the latest generation of robotized fiber systems with large fiber redundance to guarantee effective probe-to-probe sky subtraction will be the workhorses of the next decade's survey spectroscopy.