m
m pixels
The Gemini Near-Infrared Spectrometer (GNIRS) is one of the Gemini instruments assigned to the US. The instrument is being constructed by NOAO, and passed its first major milestone in the form of Conceptual Design Review, held on 22 March 1996. The next major review, Preliminary Design Review, will take place this year in October. Delivery of the GNIRS to Gemini North on Mauna Kea is currently scheduled for late 1999. The purpose of this article is to give a brief description of the instrument design, as it now stands. There may well be some changes as design work proceeds, but the basic concept should remain the same.
Science Drivers
The main design specifications for the instrument are intended to produce an instrument that will take full advantage of the unique characteristics of the Gemini telescopes: superb image quality, low emissivity, large collecting area, and the ability to carry out flexible observing.
The "baseline requirements" set by Gemini are the highest priority requirements. NOAO is obligated to meet them. These are:
m
m pixels
There are additional high-priority science "wants," of which several are included in the current design:
m cross-dispersion
Although it was not explicitly included as a requirement, a further consideration is to keep the time overheads for actual observations as low as possible. "Overhead" comprises time spent in moving mechanisms (including those external to the instrument, as well as the telescope itself), acquisition of objects, and real-time calibrations. A six-minute increase in overhead for an hour of integration is equivalent to a 10% degradation in system throughput.
Instrument Description
The capabilities of the instrument as currently designed are listed below:
Image Scales: 0.05"/pixel (long cameras) 0.15"/pixel (short cameras) Slit Length: 50" (long cameras) 100" minimum (short cameras) Wavelength Range: 0.9-5.5 um. Separate cameras for "blue" (0.9-2.5 um) and "red" (2.8-5.5 um): total of 4 Resolutions: 2000, 6000, 18,000 (long cameras) 667, 2000, 6000 (short cameras) Other: Cross-dispersion (0.9-2.4 um) Wollaston prism analyzer Rear-slit-viewing configuration On-instrument wavefront sensor Integral field upgrade path Detector: 1024 x 1024 ALADDIN detector
An optical layout is shown in Figure 1, for one of the long-focus cameras. Light enters through the dewar window and passes through a hole in a pick-off mirror to the spectrometer fore-optics. This pick-off mirror directs light from outside the general vicinity of the slit to the on-instrument wavefront sensor, which is mounted inside the GNIRS dewar for maximum stability. The hole in the mirror is sized to allow use of a slit up to 100" in length, as well as acquisition over a field tentatively set at 30" diameter.
The spectrometer fore-optics are a simple Offner system, which forms an image of the telescope secondary at its own secondary, where the Lyot stop is placed. The dewar window is in fact a weak lens, for better control of the pupil. The telescope focal plane is re-imaged by the fore-optics onto the spectrometer slit.
After the fore-optics, the light passes through the filter, slit, and then off a folding mirror onto the collimator, which is an off-axis paraboloid. The collimated light then reflects again off the fold mirror to the grating and then to the cross-disperser position. For long wavelength or long-slit work, the "cross-disperser" will be a mirror, while for cross-dispersed spectra, it will be a prism with its back side gold coated. (Note that in the final design the positions of the prism and grating may be switched; this is still under evaluation.)
The dispersed light is then reflected off another fold mirror to the
camera turret, and passes through one of four cameras. The long-focus
cameras are simple doublets, and have been folded to make them parfocal
with the short-focus cameras, which have one-third the focal length,
and contain four elements. The use of separate "red" and "blue" cameras
allows designs with very few elements that are still sufficiently
achromatic. An additional benefit is that anti-reflection coatings for
the smaller wavelength ranges are more efficient that those for the
full 1-5 
m region. One of the main considerations in the camera design
is minimizing scattered light and ghosting, to take full advantage of
the very low natural backgrounds expected between atmospheric airglow
lines below about 2.3 
m. This precludes use of field flattening
elements close to the detector.
Acquisition of objects takes place with a mirror in place of the grating (and cross-disperser); a wide slit can be used for initial acquisition.
Packaging of the instrument is still preliminary; a general view of the design is shown in Figure 2, where the main elements shown in Figure 1 are also labeled. The dewar is approximately 1.2-m diameter, and 2-m in length, not including cryocooler heads and electronics enclosures.
Total weight of the instrument, including electronics, is roughly 1600 kg. Much of this weight is due to the extremely rigid internal structure needed to meet the Gemini requirements that flexure in one hour of observation be no more than 0.1 pixel.
Upgrade Paths
As design work continues, at least two future upgrade paths will be evaluated. One is the practicality of upgrading to a larger array, which might be in the form of two 1K arrays edge-butted to form a 1K x 2K "mosaic." It is clear that the current design does not permit use of longer slits, but it appears likely that an array that is larger in the dispersion direction will work, and that greater separation of cross-dispersed orders would also be feasible.
The second upgrade is addition of an integral field capability. This is still under study, in part because the design parameters are not well specified. The basic idea behind integral field is quite simple--that instead of dispersing light from a one-dimensional slit across an array, one should instead rearrange and disperse light from a two-dimensional region, thus providing simultaneous data in a "data-cube." This approach allows efficient spectral mapping, while avoiding problems such as those due to variable seeing, variable atmospheric transmission, and differential refraction. It also has the potential of giving better spatial sampling than would be achieved just by stepping a slit across a region of sky.
The difficulty arises when one is compelled to specify the different dimensions of the data cube, subject to the constraint that all the data still need to fit onto a 1K x 1K array and then be extracted and reduced. The issue of the integral field specifications has been considered by the Gemini IR Instrument Science Working Group, but as yet no consensus has been reached. In any case, the generic approach to integral field in GNIRS does not depend much on the particular implementation. The science beam would be diverted near the slit, "processed" in some way (for example, by means of an image slicer, binary optics, or an enlarger/microlens array combination) and would then be sent back into the collimator to a point that does not require refocus of the instrument.
Jay Elias