JSTOR

3.1. Palomar 60 Inch

The Palomar 60 inch telescope is an equatorial mounted f/8.75 Ritchey‐Chrétien design (Bowen & Rule 1966). Table 1 includes the telescope prescription. CorMASS mounts directly to the instrument mounting plate at the Cassegrain focus, as shown in Figure 1. With the exception of the primary and secondary mirrors and the dewar window, all the optical elements are cooled to temperature.

TABLE 1 Palomar 60 Inch and Spectrograph Prescription

Open New Window

(231 KB)

Fig. 1.— CorMASS mounted at the Cassegrain focus of the Palomar 60 inch telescope. On the right side of the dewar is the externally mounted flip‐mirror drive assembly. In the foreground to the left is the box containing the Infrared Labs MCE‐3 electronics for clocking and array readout. The telescope visual guide camera is to the right of the dewar mounted at 45°.

Open New Window

3.2. Design Constraints

CorMASS features a compact design that was constrained by the existing dewar and the telescope focus position.

3.2.2. Telescope Focus Position The optimal telescope focus is only ∼35 mm below the instrument mounting plate (Table 1). In addition, a permanently mounted offset guide camera and filter motor assembly on the telescope instrument mounting plate constrain the outer diameter of an instrument mounted flush to the plate to be ≲10 inches.

These constraints on access made placement of a cold slit at the telescope focus difficult. The awkward focus position was good incentive to allow the beam to diverge past the telescope focus, enter the cryogenic dewar, and then reconverge to a focus onto a slit. Such a scheme would provide straightforward f/conversion between the telescope focus and slit, as well as produce a well‐defined cold pupil stop. But a reimaging scheme would further lengthen the cylindrical axis of the instrument, thus increasing expense, flexure, and the moment arm of the instrument mass that must be compensated with telescope balance.

We chose instead the more compact scheme of placing a cold slit at the telescope focus. This required a very tightly integrated mechanical design just inside the window: the dewar top plate, CaF₂ window, two radiation shield tops, and slit occupy a span of ∼35 mm measured along the optical axis. The process of f/conversion is accomplished as the beam diverges past the cold slit. The clearance between the instrument mounting plate and the telescope fork is 38 inches. The final design length of CorMASS measures 27.7 inches. Figures 2 and 3 show the design of CorMASS.

(90 KB)

Fig. 2.— The optical bench schematic. The long dimension is 14 5. This bench is easily removed from the dewar for lab testing and alignment. The flip mirror is shown in both spectrograph and slit‐viewing positions. The sides of the spectrograph camera enclosure have been removed from this drawing for display purposes.

Open New Window

(272 KB)

Fig. 3.— The optical bench removed from the dewar. Optical element mount tubes surrounding the prism are seen in the foreground and the spectrograph camera and flip‐mirror assembly is at the far end.

Open New Window

3.3. Spectrograph Optics

3.3.1. Prescription Ray tracing, optimization, and tolerancing for the CorMASS optical design were accomplished with the optical design software ZEMAX Version 6.0 (Focusoft, Tucson, Arizona). The optical layout of CorMASS is shown in Figure 4. All five lenses are spherical, and one is from stock while the remaining four are custom manufactured using in‐house test plates by Janos Technology, Inc. (Townshend, Vermont).

(53 KB)

Fig. 4.— Optical paths for the spectrograph and slit viewer. The paths split at the slit plane. The externally driven two‐position flip mirror selects either path for imaging onto the detector.

Open New Window

The main dispersing element is a coarse 40 line mm⁻¹ grating, blazed at . A prism made from the Schott dense flint SF‐10 is used in double pass for cross‐dispersion. Table 1 lists the spectrograph prescription.

After passing through the dewar window, the converging telescope beam reaches a focus at the slit. A wire electrical discharge machine was used to cut a slit into the center of a 0.008 inch thick stainless steel (SS304) disk that had been hand polished with diamond paper. Ron Witherspoon, Inc. (Campbell, California), manufactured the slit. The disk is rotated 45° about the axis of the slit length so that incoming light not transmitted through the slit is reflected into the slit‐viewing train. The slit width is oversized by to compensate for the disk rotation.

Past the slit the beam is converted from f/8.75 to ∼f/14 by an Infrasil 301‐CaF₂ air‐spaced doublet9 and is collimated by a gold‐coated, diamond‐turned, aluminum alloy (6061‐T6) off‐axis paraboloid. The beam then passes through a cold Lyot stop sized to the pupil diameter of 22.6 mm. After the stop the beam passes through the SF‐10 prism for the first cross‐dispersion pass. The prism has a 30° apex angle and is used at minimum deviation. It was fabricated by Kreischer Optics, Ltd. (McHenry, Illinois).

A prism was necessary as the cross‐dispersing element because the desired wavelength coverage exceeded a factor of 2. A double‐pass prism design was adopted because it could be implemented in combination with the grating to produce a very compact optical package. When the dispersing elements were ordered in series [prism pass 1 (cross dispersion) → grating reflection (primary dispersion) → prism pass 2 (cross dispersion)], the optical path was neatly returned into the main “two‐dimensional” plane of the optical bench (Figs. 2–4). Single‐pass prism designs led to cubical mechanical packages that could not be simply inserted into the dewar.

In addition to SF‐10, various other prism materials were considered, including AMTIR1 and ZnSe. Both have higher dispersion in the NIR than SF‐10 but were rejected because of coating difficulty and expense. The use of a material easily coated was imperative for the reduction of Fresnel reflection losses at the four vacuum‐prism interfaces encountered with a prism used in double pass.10 The dispersion of SF‐10 and the resulting order separation is discussed in § 3.3.2.

The measured internal transmission of SF‐10 (kindly provided by Dr. E. Oliva) and expected antireflection (AR) coating performance (Janos Technology, Inc.) are plotted in Figure 5. The transmission measurement was performed as described in Oliva & Gennari (1998) and includes theoretical Fresnel losses from two air‐glass interfaces (dashed line) and internal transmission through one pass of a 21.7 mm uncoated sample (dotted line). This is very close to the median length of SF‐10 traversed by the CorMASS beam in one pass (20.7 mm), so the plotted transmission, neglecting Fresnel losses, accurately reflects the expected internal transmission in our instrument. Shortward of , the glass is nearly transparent. We assume absorption features longward of are due to water/OH in the glass. Combining the expected reflection losses through two air‐glass interfaces after AR coating (long‐dashed line) with the internal transmission gives the total expected transmission through one prism pass for CorMASS (solid line). While the transmission losses were unfortunate, they were accepted because the primary spectral features in K band for L and T dwarfs, the CO band head at and the feature, respectively, were deemed less important and redundant compared to the źJH band features.

(34 KB)

Fig. 5.— Measured internal transmission of SF‐10 (21.7 mm thickness) (dotted line) including theoretical reflection losses from two air‐glass interfaces (dashed line). Square boxes represent absorption prediction derived from Schott data sheet for uncoated material. Expected AR coating performance for two air‐glass interfaces (long‐dashed line) is applied to the measured internal transmission to estimate total transmission for one prism pass in CorMASS (solid line).

Open New Window

Following the first pass through the prism the beam is diffracted by the grating. Richardson Grating Labs (Rochester, New York) mounted an epoxy replica of a stock master grating onto an aluminum alloy (6061‐T6) substrate provided by us. The gold‐coated grating, with 40 line mm⁻¹ groove density and blazed at 5 5, is used with a turn angle of 21 5 to match the angular beam deviation of the off‐axis paraboloid for optical layout symmetry. The first‐order blaze wavelength ( ) is . The detector images the second through ninth orders, with coverage from to . We designed the spectrograph to use orders 2–5 and were fortunate that orders 6–7 are in good optical focus as well. In practice, only orders 2–6 have useful throughput. (Spectrograph throughput is discussed in § 6.)

Inadvertently the shortward edge of order 2 and longward edge of order 3 are cut off by the array edges, resulting in a gap in spectral coverage from 2.014 to . Evidently the final spectrograph camera focal length is too long, likely owing to errors in our assumed cryogenic indices of refraction.

Grating efficiency is well described by scalar theory if , where d is the groove spacing ( for CorMASS). Therefore, polarization effects are very small (Loewen et al. 1977). Theory also predicts ≳50% efficiency from to in first order. At higher orders this efficiency will be maintained because the grating is used even farther into the scalar domain as decreases (Loewen et al. 1977). A check of our chosen grating parameters in a commercial software program for grating efficiency analysis (PC Grate, Optometrics, Ayer, Massachusetts) yields estimated peak efficiencies of 90% for orders 2–5.

After diffraction by the grating, the beam passes through the prism for a second time. Finally, the beam is focused onto the detector by an f/5.5 camera which delivers a final plate scale of 1 pixel⁻¹ to the square NICMOS3 pixels. The camera features three elements in a doublet‐singlet format. Following a BaF₂‐Infrasil air‐spaced doublet, there is a space of ∼5 inches before the converging beam passes through the final Infrasil lens and focuses on the detector. The final lens and detector are separated by ∼1 inch.

The ratio of the telescope central obscuration diameter to primary diameter is . For a telescope with the encircled energy fraction (EE) within the first Airy dark ring is 0.65 (Schroeder 1987). We used this EE benchmark to judge the image quality of the spectrograph design. Shortward of all wavelengths image within 1 pixel (Fig. 6). This applies for the entire slit FOV ( ).

(39 KB)

Fig. 6.— The 65% encircled energy diameter vs. wavelength for the spectrograph (entire slit FOV). The approximate diffraction limit (dashed line) and the pixel size (dotted line) are also shown.

Open New Window

3.3.2. Order Format The SF‐10 prism in CorMASS leads to an order format similar to that found when a first‐order grating is used as the cross‐disperser;11 the order separation of CorMASS increases with increasing wavelength (Fig. 7), whereas prism cross‐dispersers typically give larger order separation for decreasing wavelength. The CorMASS format occurs because the dispersion of SF‐10 in the NIR does not follow the usual dispersion law for transparent glasses in the optical, . Instead of a monotonically decreasing dispersion with increasing wavelength, SF‐10 has a dispersion minimum at ∼1.7 μm (Fig. 8) at room temperature. This leads to theoretical relative order separations also shown in Figure 8. Both plots compare SF‐10 and ZnSe. SF‐10 has a similar dispersion function at 77 K as at room temperature.

(68 KB)

Fig. 7.— Slit viewer and spectrograph images. (Left) Jupiter fills the 35 slit viewer FOV, with Io just below the slit. Row and column directions are indicated. (Right) The spectral image displays a Wolf‐Rayet star near the middle of the slit. The second order is at the top, sixth order at the bottom. The thermal emission of the atmosphere increases rapidly in the second order then abruptly cuts off at ∼2.5 μm. Sky lines dominate the third order (H band). NICMOS3 shading anomaly is seen at the bottom of the quadrants prior to correction. Order numbers and direction of increasing wavelength are indicated.

Open New Window

Fig. 8a – 8b

(1 of 2 images)

(70 KB)

Fig. 8.— (a) Dispersion ( ) vs. λ and (b) relative order separation vs. λ for Schott dense flint SF‐10 and ZnSe at room temperature. SF‐10 is used for the cross‐dispersing double‐pass prism.

Open New Window

3.4. Slit Viewer Optics

Complementing the spectrograph optical train is a slit‐viewing mode (Fig. 4). Light not transmitted through the slit is reflected by the polished slit substrate into the secondary slit‐viewing train. This allows viewing a image of the reflective slit for rapidly steering science targets into the slit. A remotely controlled drive motor outside the dewar turns an internal flip mirror between the slit‐viewing and spectrographic positions in ∼10 s. The flip mirror is located between the doublet and singlet of the spectrograph camera. The slit‐viewing optical prescription is listed in Table 2. Since the slit‐viewing mode is not intended for photometric imaging, all lenses in the slit‐viewing train are from manufacturer’s stock to reduce material costs.

TABLE 2 Slit Viewer Prescription

Open New Window

Light reflected from the slit is collimated by a CaF₂ lens and passes through a Lyot stop sized to the pupil diameter of 5.1 mm. The collimated beam is then steered by two gold‐coated fold mirrors to the spectrograph camera assembly. The beam also passes through a filter mounted between the fold mirrors (tilted 5° away from the spectrograph elements to reduce light contamination). The slit‐viewing filter can be changed by opening the dewar. Finally, the beam is refocused onto the detector by the combined action of a CaF₂ lens and the final Infrasil lens of the spectrograph camera (Fig. 4). The gold‐coated flip mirror steers the light from the CaF₂ lens toward the final Infrasil lens and detector. The plate scale of the slit viewer is 0 27 pixel⁻¹, corresponding to an f/18.85 beam. This plate scale was chosen to give an FOV (35 ) large enough to aid the identification of targets in confused fields (Fig. 7). The slit viewer design gives 65% encircled energy diameters of ≤3 pixels on axis and ≤4 pixels at the edge of the FOV.

Light leakage from the slit‐viewing train when in spectroscopy mode is minimal because the flip‐mirror mechanism blocks the slit‐viewing beam just past the second slit‐viewing lens with a shaped baffle plate when not in use (Fig. 2). We gave extra care to minimize this source of stray light in such close proximity to the detector.

4.1. Mechanical Design

AutoCAD Version 14.0 and Mechanical Desktop 2.0 (Autodesk, San Rafael, California) were used for the mechanical design. The optics were designed assuming all elements inside the dewar are at 77 K. Since the optical mounts are machined at room temperature, distances between elements required by the optical prescription are adjusted for the linear expansion of materials (Table 3) prior to the mechanical design and fabrication. Thus, upon cooling the elements are correctly positioned.

TABLE 3 Coefficients of Thermal Expansion

Open New Window

All optical mounts were fabricated from the aluminum alloy 6061‐T6, the same material used for the off‐axis paraboloid and grating substrate. The mounts are attached to a “T” cross‐section aluminum “optical bench” extending from the dewar’s cold plate. The use of only one material for all mounts reduced coefficient of thermal expansion effects and eased the difficulty of achieving good optical alignment at 77 K. The aluminum optical bench, off‐axis paraboloid, and grating substrate were thermally cycled between room temperature and 77 K multiple times prior to final finishing to stabilize them.

The prism is constrained through a system of six aluminum pads that form a semikinematic mount (Fig. 9). Three spring plungers oppose the pads in three dimensions, and each exerts 1.0–1.6 pounds force to constrain the ∼0.7 pound prism. Between each pad and the prism material is an indium shim to prevent abrasion (Fig. 9).

Fig. 9a – 9b

(1 of 2 images)

(63 KB)

Fig. 9.— (a) The prism cage constrains the prism in a semikinematic mount and provides stray light control. (b) The prism holding scheme includes three spring plungers that oppose six monolithic pads in three dimensions to secure the prism and compensate for thermal expansion effects.

Open New Window

To accommodate differing thermal contractions, all lenses are mounted in lens barrels with oversized inside diameters to prevent crushing. Retaining rings compress spring wave washers against the outer annulus of lens faces to accommodate differences in axial contraction during cooldowns.

The flip mirror is driven by an externally mounted 30 VDC motor. A shaft directly couples the external motor to the flip mirror through an O‐ring rotary seal at the dewar wall. The O‐ring seal and a flexible coupling inside the dewar make up for small positional and axial changes upon cooling the instrument. Mirror position is controlled by a logic circuit involving a remotely operated toggle switch at the observer’s station and limit switches engaged by the motor shaft outside of the dewar. Placement of the limit switches outside the dewar allows adjustment of the mirror throw while the instrument is cold. The double‐shielded dewar uses only ∼1 liter of per night of observing.

4.2. Stray Light Reduction

All appropriate surfaces are painted black for stray light reduction. After using one primer coat of the two‐part 9924 Wash Primer, black polyurethane Aeroglaze Z‐306 paint (Lord Corporation, Erie, Pennsylvania) mixed with carbon black powder was applied in two or more coatings to all exposed aluminum surfaces (Hunter et al. 1996).

The optical element mounts were designed to minimize stray light. Lenses are mounted within cylindrical aluminum tubes, as mentioned above. Particular attention was devoted to reducing stray light adjacent to the prism. Since the reflective slit is close to the prism, both optical paths past the slit are well enclosed in tubes. The prism itself is enclosed in a mount that exposes only the entrance and exit faces to the optical beam. Another critical area was that near the detector. To reduce stray light, the final spectrograph camera and flip‐mirror assembly were placed within a rectangular enclosure that extends nearly to the cold plate to prevent the direct illumination of the detector by stray light from the shield walls (Fig. 3). All transmissive elements in the spectrograph train were AR coated.

4.3. Alignment

Alignment of the optical trains was completed in two parts—internal element alignment and alignment with the telescope secondary. First, alignment within the instrument was conducted in the lab with an HeNe laser. As discussed above, because all element mounts were made from aluminum, the relative alignment of the optical elements was preserved upon cooling from room temperature to 77 K. A telescope simulator was used to reimage a neon‐illuminated point source onto the slit plane to check focus. The slit‐viewing collimating lens and spectrograph f/conversion lens positions were adjusted to make the slit viewer field and the spectrum confocal with the slit.

Second, the independent optical paths of the spectrograph and slit viewer were co‐aligned with the telescope secondary through an iterative process. Using a gimballed sighting scope mounted on the telescope at the same position at which CorMASS normally mounts, the scope was centered on the center of the secondary. The scope (with its directional alignment preserved) was transferred to a Cassegrain mounting plate simulator in the lab and sighted onto the paper disk the diameter of the secondary and mounted at the distance of the secondary. CorMASS was then mounted onto the simulator, and a neon lamp was used to trace the f/cone of both optical paths at the distance of a paper disk. Initially, the spectrograph f/cone was found to be aligned within 0 5 of the secondary (centers within 1 inch at the slit‐secondary distance of 135 inches), but the slit‐viewing f/cone was found to be displaced 4 inches. The slit‐viewing train elements were realigned, and a subsequent test showed the alignment to have improved to within 0 5.

Peak‐up of both trains relative to the secondary can be fine‐tuned by adjusting the position of four mounting blocks that attach the instrument to the Cassegrain mounting plate.

7.1. Data Reduction

CorMASS first‐light observations occurred during 1999 August 22–25, with seven additional runs through 2000 August.

Single‐frame integration times of 100–300 s are typically taken with telescope nods of ∼4 5–7 along the slit between each integration. Frame integration times are reduced when airglow and/or atmospheric water content are judged to be highly variable to maintain effective sky subtraction.

Dome flats are taken before the start of each night’s observing. Dome illumination for the flats is provided by a 10 W QTH (quartz, tungsten, and halogen) lamp mounted above the telescope secondary.

Spectral observations of calibrator stars for atmospheric correction are taken at air masses within ±0.1 air mass of the target observations. A stars from the Elias standards (Elias et al. 1982) and bright late‐F and G stars have been used as calibrators.

Following bad‐pixel correction, the flat‐field and flux calibration spectral images are corrected for NICMOS3 shading. This is a noiseless exponential change in pixel bias based upon position, time since last pixel read, and detector temperature. The changing temperature of the detector readout amplifiers during readout causes an exponentially varying bias in the direction orthogonal to the pixel clocking or along the columns (Böker et al. 2000). The columns are approximately perpendicular to the orders (Fig. 7). To remove this artifact, a quadratic is fit to a row‐by‐row clipped median of the top quarter of the array, the portion unused by the spectrograph. The fit is extrapolated to 128 rows for the top two quadrants, duplicated for the bottom two quadrants, and subtracted row by row from all pixels. The quadratic fit agrees well with row medians of “quasi‐dark” images taken with the flip mirror halfway between spectrograph and slit‐viewing mode.

Spectra are reduced using standard routines in IRAF.12 Images are flat‐fielded using a pixel responsivity solution derived from the APFLATTEN task using dome flats from all nights of the run summed together. Nodded image pairs for science objects are subtracted against each other to remove sky background to first order and shading effects. The shading correction discussed above is not applied to science object images because the subtraction of adjacent nodded images removes the effect to a comparable noise level as the quadratic fit method.

The APALL task is then used for spectral extraction. Wavelength calibration is accomplished by using line identifications from spectral observations of planetary nebulae, e.g., NGC 7027 and NGC 6210. A third‐order polynomial is fit to the identified lines to derive a dispersion solution across the orders. The fit residual rms is ≤ pixel (2.25 Å) across the wavelength coverage. The use of OH airglow lines was rejected because at the resolution of CorMASS ( ), many airglow lines are blended together, especially in the H band, making line identifications problematic.

Telluric absorption effects are removed by dividing by the reduced spectrum of a standard observed at similar air mass and hand corrected for stellar H Paschen and Brackett recombination lines as required. The ratio is then multiplied by a blackbody with a temperature equal to the effective temperature of the stellar type of the calibrator observed (Tokunaga 2000). Finally, all observations for each order are weighted by the spectra mean and combined with the SCOMBINE task using a trimmed average. Orders are stitched together with SCOMBINE after hand deletion of noisy data at the ends of the orders.

7.2. Spectra

A spectrum of the infrared bright planetary nebula NGC 7027 is shown in Figure 11. The plentiful emission lines of planetary nebulae, primarily H and He lines, are used for wavelength calibration. The wavelength coverage and efficiency of CorMASS are well suited for the study of numerous types of objects including quasars, novae, and emission‐line objects.

(60 KB)

Fig. 11.— CorMASS spectra of NGC 7027. This well‐studied planetary nebula was observed 1999 October 24 for s. A total of 38 distinct emission lines across six orders were identified for use within IRAF to derive a dispersion solution for the wavelength calibration of program objects. This plot has been clipped to enhance fainter lines.

Open New Window

Spectra of the low‐mass objects VB 10 (M8 V) and 2MASSW J1507476−162738 (L5 V, hereafter 2M1507) and the T dwarf 2MASSI J0559191−140448 (hereafter 2M0559) are presented in Figure 12. A full spectral sequence and NIR L dwarf features will be discussed in another paper (J. C. Wilson, M. F. Skrutskie, J. E. Gizis, D. Monet, J. D. Kirkpatrick, & J. R. Houck 2001, in preparation).

(40 KB)

Fig. 12.— CorMASS źJHK spectra of three low‐mass objects: M8 V (VB 10), L5 V (2MASSW J1507476−162738, hereafter J1507), and a T dwarf (2MASSI J0559191−140448, hereafter J0559). VB 10 ( ; Leggett 1992) was observed 1999 August 22 for s. J1507 ( ; Kirkpatrick et al. 2000) was observed 2000 April 20 for s. J0559 ( ; Burgasser et al. 2000) was observed 1999 October 24 for s. Data between bands was not usable owing to poor atmospheric subtraction.

Open New Window

CorMASS spectroscopically confirmed the classification of 2M0559 ( ) during the instrument’s second run (Burgasser et al. 2000). 2M0559 is 0.4 mag brighter than Gl 229B and more than 1 mag brighter than the other field T dwarfs discovered to date. This object’s brightness will aid broad‐wavelength and high‐resolution study of its spectral class.

While no methane combination and overtone absorption features at 1.6 and (the features that define the T dwarf class) are evident in 2M1507, weak fundamental methane absorption has been discovered at in 2M1507 and in an L7.5 (Noll et al. 2000). This suggests that the transition from CO to as the primary carbon‐bearing molecule as effective temperatures decrease commences as early as L5 V.