P.I.: Dr. Ian McLean, UCLA

August 19, 1999

NIRSPEC, the near-infrared cryogenic echelle spectrograph for Keck II, will be available for general use in Semester 2000A. This instrument, produced by UCLA in collaboration with UCB and CARA, achieved "first light" on April 25 and completed commissioning trials on August 18, 1999. A summary of NIRSPEC's properties and performance is given below. There is a draft User's Manual available at CARA HQ. We are still working on making the Manual and the IDL-based Echelle Format Simulator available over the Web. Prospective users should consult the "instruments" section of the Keck Web page from time-to-time for more details and updates, or contact Tom Bida, the Instrument Specialist for NIRSPEC.


NIRSPEC is optimized as a high-resolution spectrograph for the wavelength region from 0.95-5.5 µm. The resolving power (R=/) is 25,000 (12 km/s) for a slit width of =0.43" (3 pixels). Alternatively, the R product is 10,800. A special Low-Resolution mode of R=2,500 is also provided. NIRSPEC has 3 detectors: a standard CCD camera for acquisition and offset guiding, a 1-2.5 µm near-infrared "slit-viewing" camera - the SCAM, and the primary 1024 x 1024 IR array for the spectrograph itself.


NIRSPEC resides on kinematic mounts on the Right Nasmyth (RNAS) platform of the Keck II telescope and operates at the f/15 focus with the tertiary mirror in place. The Keck focal plane falls about one inch in front of the entrance window of the vacuum enclosure (or dewar). Optically, the instrument consists of a "front end" which collimates the diverging f/15 beam to produce a pupil image approximately 26 mm in diameter and provide a convenient location for a cold stop. This section is followed by re-imaging optics which convert the beam from f/15 to f/10 (2.06"/mm) at the slit plane. After the slit, the emergent f/10 beam is collimated with an off-axis parabolic mirror to produce a 120 mm diameter pupil on the echelle grating. Diffracted light from the echelle is then cross-dispersed by another grating at right angles to the first, and the beam is collected by a special f/3 camera in the "back end" and then focused onto a large-format infrared array detector. The design employs all-reflecting surfaces. Each mirror is diamond-machined and post-polished on a nickel-aluminum substrate. Both gratings also use aluminum substrates. The mirrors are silver-coated, whereas the gratings and slit substrates are gold-coated. A large, custom-designed vacuum chamber - the Big Red Box - encloses all the optics which are cooled to about 60 K, with closed-cycle refrigerators, to achieve a very low background (about 0.2 electrons/s/pixel) within the instrument. Refilling with liquid nitrogen is not required in normal operation. The vacuum-cryogenic chamber, and all the necessary electronics, is supported by a steel handling frame which can be moved on rails by a built-in motorized handling cart. NIRSPEC is controlled from a Graphical User interface which includes a powerful Echelle Format Simulator (EFS).

The spectrograph employs a single echelle grating in Quasi-Littrow (QLM) mode with the out-of-plane angle, = 5°. This approach provides significantly higher efficiency than the Near Littrow mode with the same opening angle, with the minor penalty of slit images which are slightly tilted. With tanB = 2, where B = 63.5° is the blaze angle, and a (cryogenic) groove density of T= 1/d = 23.29 lines/mm, the product of blaze (or central) wavelength and order number is given by

mB = d (sini + sino)cos = 2dsinBcos = 76.56 microns.

At the short-wavelength limit, a central wavelength of = 0.957 µm occurs in order m = 80, but near the detector cut-off wavelength, the central wavelength of = 5.47 µm occurs in order m = 14. The free spectral range (FSR) in each order is /m and to fit one FSR onto a detector with N pixels at 3 pixels per resolution element we need mfsr = 3R/N. For R=25,000 and a 1024 pixel array this implies mfsr = 73 and the corresponding central wavelength is 1.045 µm; the FSR is therefore 0.014 µm or 140 A and the dispersion at this wavelength is 0.139 A/pixel. At longer wavelengths the FSR is much larger than the array and therefore multiple grating settings are needed for complete coverage of a given band. Since the "extent" of a free spectral range in microns scales with the wavelength squared, but the resolving power is constant and therefore the "size" of a resolution element () in microns scales with wavelength, then the number of detector "settings" needed to cover the width of the echellogram scales only as . For example, at = 5.225 µm (or 5 x 1.045 µm) the number of non-overlapping settings would be 52/5 = 5.

For cross-dispersion, the grating used has a smaller blaze angle, higher groove density and zero out-of-plane angle. Since the geometry of the optical layout requires a difference of 50° between the input and output beams, and the condition for maximum efficiency is met when |i - B| = |o -B|, then we find that i ~ 35°, o ~ -15°, and B ~ 10°. NIRSPEC's cross-disperser grating (or CDG) has a groove density of 75.75 lines/mm (at the cryogenic operating temperature). Using these values in the grating equation yields

mccen = 4.155 microns.

The CDG is used in low order (mc = 1 - 4) in conjunction with order-sorting filters located in a filter wheel near the pupil image formed by the front-end optics. Order separation in pixels is 14.25B2mc which gives about 62 pixels at 1.045 µm in 4th order, or about 11.8".

The echelle is mounted back-to-back with a flat mirror. Replacing the echelle grating with the flat mirror yields a convenient low resolution ("low-res") mode from the CDG alone with R = 2,500 (120 km/s) for a 2 pixel wide slit. The dispersion is approximately 2 A/pixel at 1 µm and scales with , and the free spectral range always exceeds the bandwidth of the order-sorting filters.


It should be noted that all filters are common to the spectrograph and imaging channels. Seven custom-designed filters, referred to as NIRSPEC-1 through NIRSPEC-7, provide over-lapping wavelength coverage from 0.95 - 2.6 µm, and there is a KL, L, M and Mwide for 3.0-5.5 µm. NIRSPEC 3 and NIRSPEC 5 correspond roughly to the J and H bands. NIRSPEC 6 is a broad H+K filter centered at 1.925 µm with a bandwidth of 0.75 µm. Standard K and K´ filters are also included. The very broad KL filter allows the SCAM to be used for guiding/imaging in the K band while spectroscopy is performed in the L band; this filter reaches the Brackett alpha line at 4.05 µm. Several one percent wide "narrow-band" filters are also provided, such as Br at 2.165 µm and the H2 S1 line at 2.122µm. Plots of the spectral transmission curves of all these filters are available on the web page and from within the Echelle Format Simulator (EFS); the EFS is described later.


A selection of polished, laser-cut, air-spaced slits are mounted in a wheel in the re-imaged focal plane. Each slit is tilted by about 12° to enable reflected light to reach the independent infrared slit-viewing camera (the SCAM).

Slits of 1, 2, 3, 4 and 5 pixels are available with 12" length, and widths of 2, 3, 4 and 5 pixels are provided at 24" length. Shorter slits are needed to avoid order overlap at the very shortest wavelengths. The order overlap can be inspected with the EFS. About 10 orders are captured at 1 µm but only about three at 4 µm due to the larger order separation at longer wavelengths. In the Low-Res mode the available slit widths are 2, 3 and 4 pixels and all of the low-res slits are 42" long. There is no imaging mode which does not contain a slit, but a "box-9" dithered imaging pattern using the short, narrow slit (1 x 60 pixels) is extremely effective in removing the effect of the slit and going deep.

All slits are rotated relative to the rows and columns of the slit-viewing camera. These angles are built in to the software for nod commands. In the echelle mode, these built-in slit rotations, as well as a built-in detector rotation, enable the echelle orders and slit images to be aligned along the rows and columns of the InSb array.

The Three-Mirror Anastigmat (TMA) camera in the spectrograph has different focal lengths along the dispersion and spatial (slit height) directions. In the Echelle mode, this circumstance yields an image scale of 0.144"/pixel in the dispersion direction and 0.193"/pixel along the spatial direction on the infrared array. Note however, that in the Low-Res mode these scales are reversed. Three pixels in echelle mode is 0.432" whereas 2 pixels in low-res mode gives 0.386". The image scale on the SCAM is 0.183"/pixel and on the CCD camera it is 0.21"/pixel.


A unique feature of NIRSPEC is its cryogenic optical Image Rotator which enables any arbitrary position angle (PA) to be aligned with the fixed slits or with the rows and columns of the SCAM. The image rotator can also be placed in "tracking" mode to automatically compensate for field rotation. Since the Image Rotator precedes the SCAM, field rotation is removed from the SCAM images. The Image Rotator can be used in Position Angle mode or in Physical mode. A graphical user interface is used to set up the Image rotator and give a "preview" of the physical position and direction of rotation of the rotator. We have noticed some slippage of the rotator which may be getting worse with time, due presumably to wear and increased friction of the bearing. It is advisable to re-initialize the Image Rotator fairly frequently (between exposures) to minimize build up of field rotation.


NIRSPEC is operated remotely using a suite of Graphical User Interfaces (GUI's). These GUI's provide the following:

a Status and Instrument Control panel - which includes a cartoon of the light path,

a pair of Exposure Control panels - one for each infrared detector,

separate re-sizeable IDL-based Quick Look display windows for the SCAM and Spec cameras,

a graphical display for the Image Rotator,

an IDL-based Echelle Format Simulator (or EFS).

The EFS is very powerful and provides a means of generating scripts AND controlling both the instrument set-up and the spectral exposures. The progress of the current integration in either the spectrograph or the SCAM is shown on the Exposure Control panels together with the filename being written. All files are written in FITS format with all available keywords included in the image headers.

Set-up of the instrument can be done "one mechanism at a time" using the buttons on the Status & Control screen. For example, one can select the filter, or slit from a "drop-down" menu selection, or set the grating angles, or control the Calibration Unit. Alternatively, the entire configuration plus exposure times can be set from within the Echelle Format Simulator. No mechanism move takes longer than about 30 seconds. The EFS draws an accurate representation of the expected echellogram (or low-res spectrum) and the edges of the detector for each selected order-sorting filter. As each filter is selected the box representing the detector first appears in the central position. By "dragging" the box with the mouse over the echellogram, the effect of "scanning" the gratings is reproduced. Once the optimum location is found, the grating angles can be read off, typed into the Status and Instrument Control panel and activated with the "set" buttons. Alternatively, the EFS "go" button can be used to drive all the mechanisms (slit, filter, echelle, and CDG) in parallel to the selected set-up. Note that the EFS provides an easy and powerful way to know on which pixel a given wavelength "should" occur.

Calibration arcs and flat fields can also be pre-programmed using the EFS. It is strongly advised that a set of arc lines be taken at each grating setting since the mechanisms cannot be relied upon to repeat with sufficient accuracy. Four arc lamps (neon, argon, krypton and xenon) are provided together with a quartz-halogen lamp for flat fields. To perform a calibration manually without using the EFS, the Status and Instrument Control GUI must be used to move the Cal Unit fold-mirror in front of the dewar window and switch on the selected lamps. Exposure times and co-adds must then be entered into the Spec Exposure Control window. In low-res mode, exposures of about 0.25 s are sufficient for the arcs, with about 10x longer for high-res mode. The flat field lamp requires about 1-5 s depending on wavelength in low-res mode.


Guiding can be performed either with the standard PXL CCD camera or with the SCAM. The CCD camera is controlled by the Observing Assistant, but control of the SCAM is "shared". When the SCAM is used for guiding then control must be given up by the user so that the telescope operator can take over. Sky-subtracted SCAM images will then appear on the xguide camera display instead of the CCD images. Infrared guiding on slit-spillage is particularly useful during twilight observing.

Acquisition can be performed using either camera. Two pointing origins have been defined for NIRSPEC. If the CCD camera is used to define the object then the telescope operator will use the reference point known as REFA (top center of annulus when Position Angle of rotator is zero) whereas if the SCAM is used then a point near the center of the array called REF is used. Note that the point REF does not fall on the slit. In principle, any position angle of the Image Rotator can be used to acquire a source. Once a SCAM image is obtained and displayed with the Quick Look software, tools are available from pull-down menus to move the telescope to place the object on the slit. If the object is faint, a "box 9" of nine slightly displaced SCAM images can be taken and median-filtered to produce a "deep" image. This entire option, including the on-line reduction, is built into a single keystroke or button on the SCAM GUI.

If the SCAM is not the guider, then it can be used at any time during a long spectroscopic exposure to "check" centering or obtain deep images.


In the spectrograph section the detector is a Raytheon/SBRC 1024 x 1024 (ALADDIN) array of indium antimonide (InSb) with 27 µm pixels and a dark current of 0.2 electrons/s/pixel. A conversion gain of 5 electrons/data number (DN) is used with the InSb array and the "safe" limit on well-depth (with a 300 mV reverse bias) to maintain good linearity is 100,000 electrons or about 20,000 DN. Three sampling modes are available: single sampling, correlated double sampling (CDS) and multiple correlated double sampling (MCDS) - also known as Fowler sampling. The shortest exposure time is 0.25 seconds. In normal operation with exposures longer than about 5 seconds, the default mode is 16 multiple reads (Fowler 16) which yields a read noise of about 25-30 electrons rms with the currently available array; this is significantly worse than expected.

Assuming that the read noise is about 33 electrons then the sky background signal must exceed 1000 electrons (200 DN) per pixel to achieve background-limited performance. At non-thermal wavelengths below 2 microns, the sky flux between the OH lines is about 0.33 electrons/s/A. Strictly, on-chip exposures > 1200 s are needed in low-res mode to achieve background-limited performance. This is not always practical and may generate unwanted artifacts from the chip; 600-900 s may work better. The high-resolution echelle mode is read-noise limited with the current InSb array for any practical exposure time at all wavelengths less than 2.2 microns. In the thermal IR both modes are background-limited. A signal-to-noise ratio of about 10 (read-noise limited) can be obtained in the low-res mode for an on-chip integration time of 600 s at H = 16.5.

In the SCAM section, a low-noise Rockwell 256 x 256 (PICNIC) array of HgCdTe detectors with a cut-off wavelength of 2.5 microns provides a field of view of about 46" x 46" and a scale of 0.183"/ pixel. At the operating temperature of 60 K, dark current in this detector is negligible but the quantum efficiency is much less than the InSb detector. Images with full width half maximum (FWHM) values of 2-3 pixels (0.37" - 0.55") were seen routinely during commissioning. Using a conversion gain of 4 electrons/DN with the HgCdTe array, the safe limit to maintain good linearity is 120,000 electrons or about 30,000 DN. Again, three sampling modes are available: single sampling, correlated double sampling (CDS) and multiple correlated double sampling (MCDS). The shortest exposure time is 0.10 seconds. The detector readout noise is typically about 10 electrons (CDS) and therefore this camera is background-limited when the flux exceeds 100 electrons (25 DN) per pixel. Since the zeropoint corresponding to 1 DN/s is about 24.7 magnitudes in the J band with SCAM and the J background is typically about 15.4 magnitudes per square arc second, or about 158 DN/s, this camera is almost always background-limited.

The third camera in the system is a standard PXL CCD camera from Photometrics Ltd., which contains a 1024 x 1024 pixel SITe CCD giving a plate scale of 0.2"/pixel. This camera, which is mounted outside the main dewar, is used to view the entire annular field surrounding the SCAM field from an inner diameter of about 1 arc minute out to a diameter of about 3.5 minutes of arc. The total sky area available in this annular field measures approximately 7.5 square minutes of arc and guide stars as faint as R = 20 can be used; the CCD camera is fitted with a fixed R-band filter. Since this camera is in front of the Image Rotator there is no compensation for field rotation, but software takes this into account.

The SCAM camera is quite sensitive. For example, the 1 sigma 1 second limit in the H band is about 17.5. The sky background in the K filter is about 13.4 magnitudes per square arc second, about 0.2 magnitudes better at K´, and a K´=17.8 gravitationally-lensed galaxy at a redshift of z = 2.7 was easily detected in the difference of two 20 second exposures under good (3 pixel) seeing conditions.