Chapter 4

Chapter 4, The Point Design

Section 4.7.3: Infrared Echelle Spectrographs


Origins of Planetary Systems

One primary science driver for high-resolution spectroscopy in the near- and mid-infrared spectral region is the study of the origins of planetary systems. The goal is to understand where and when planets form and to infer the planetary architectures through the observation of gaps within the protoplanetary system disk surrounding the young central star.

In order to do this, spectroscopy of the protoplanetary disks is needed with a resolving power of 100,000. There are two optimal spectral regions to do this: one is located near 5 microns, and the other is near 17 microns. The signature of the observation reveals itself as dual emission peaks with a separation equal to the orbital velocity corresponding to the radial distance of the gap from the star.

Key requirements for such an instrument, besides the high spectral resolution, are on-axis, moderately high Strehl AO images of the stellar system and low emissivity. The combination of high Strehl and low emissivity enhances the detection of the protoplanetary disk by suppressing background contamination of the sky and telescope. The diffraction-limited images also make it easier to achieve the spectral resolution required by shrinking the input slit and beam diameter of the spectrograph. INSTRUMENT DESIGN ISSUES

We present here preliminary design studies for two high-resolution IR spectrographs for the GSMT. The task was to design a 1.5 - 20 micron diffraction-limited spectrograph. The wavelength span is very large, more than an order of magnitude in wavelength, and the technology required changes considerably across this spectral range. This, combined with the large change in the size of the Airy disk, requires that the spectrograph be divided into three separate spectrograph modules, 1.5 - 5 microns, 8 - 14 microns, and 16 - 20 microns. The wavelength division has been done in regions where the telluric spectrum is opaque for all ground-based observers. We will describe here design concepts for a near-IR spectrograph (1 - 5 microns) and a mid-IR spectrograph (16 - 20 microns). The 8 - 14 micron spectrograph could be very similar to the 16 - 20 micron spectrograph, and might utilize some of the same components. It is unlikely that the 8 - 14 micron regime could be identical to the 16 - 20 micron instrument due to the factor of two difference required to properly sample the diffraction-limited input.

The scientific goals require high spectral resolution (as mentioned above in the specific case of protoplanetary disks). Resolution of at least 50,000 is also desirable for many types of stellar studies. For work on the ISM (Interstellar Medium), the intrinsic line widths are on the order of a few km s-1 requiring resolution of at least 105. Resolution on the order of 105 is also suitable for abundance studies. We feel that resolution greatly exceeding 105 results in the requirement of an impractically large instrument, and also restricts the range of applications of the spectrograph. Thus, for this design study, a resolution goal of 100,000 to 120,000 was set. This is in accord with the highest resolution for a number of existing facility spectrographs, e.g., the HET (Hobby-Eberly Telescope) echelle.

The current generation of IR spectrographs do not provide full wavelength coverage of extended intervals. However, this feature is commonly provided in the optical, and we believe that it should be a key element for the next generation of IR spectrographs. Full wavelength coverage will allow the IR to be fully exploited for those objects with large redshift or with complex broad and narrow line spectra. Our goal was to produce designs with wavelength coverage limited by detector physics or by large changes in observing conditions (the turn-on of thermal background radiation in the 2.5 - 3 micron region, for example), rather than by mechanical/optical limitations. Technical Issues

We encountered technical challenges in this investigation in the areas of optical components, thermal design, and optical design.

The major issue in optical components is the selection of gratings. The echelle gratings required can be large and must be coarsely ruled. We believe that diamond machining provides a solution to this problem. As the wavelength increases, the scale of the instrument becomes larger, and a larger format grating is required to achieve the same resolving power. This is due to the increasing size of the diffraction-limited image as wavelength increases.

In the area of thermal design, we found that although the 1.5 - 5 micron spectrograph NIrES is on the scale of current generation instrumentation, the 16 - 20 micron spectrograph MIHDAS (mid- infrared high dispersion AO-fed spectrograph) is room size. Because of the need to eliminate thermal radiation arising in the instrument, these instruments must be cooled. The NIrES need only be cooled to ~ 60 K, within the range easily reached by closed cycle refrigerators. However, MIHDAS must be cooled to ~ 10 to 20 K. Although room-sized instrumentation has been cooled to LHe temperature in physics laboratories (e.g., Fermilab facilities), this is a new challenge for astronomical instrumentation, particularly for a remote mountaintop environment.

In the area of optical design, the camera designs for the spectrographs proved difficult. The requirement of cross dispersed orders filling fairly large detector arrays required cameras with large fields of view. For NIrES, we present a strawman lens design that largely accomplishes our goals and that has excellent image quality. We were not able to arrive at a final solution for MIHDAS. The problem here is quite difficult due to the large size of the collimated beam and the requirement for an all-reflective camera. For that instrument, we present a four-mirror design that appears promising. Detector Options

The wavelength range spans three possible zones of detector technology. In the 1 - 2.5 micron region, HgCdTe arrays of 2 K x 2 K format are becoming available. Rockwell is also producing a HgCdTe detector that can span the full 1 - 5 micron region. InSb detectors, produced by Raytheon, are also viable in the 1 - 5 micron regime, and are (or will be) also available in 2 K x 2 K format and two-side buttable. Longer wavelength detectors are typically doped silicon arrays and are available with a 1 K x 1 K format and three-side buttability. Typical pixel sizes for IR arrays are in the 25 micron range.

Within the next decade, substantial improvements should be expected in the performance and noise characteristics of such IR arrays. Grating Options

The selection of the main dispersive element proved the critical feature in the design of these spectrographs. We considered three options: existing echelle (replica) gratings, immersed gratings, and machined gratings. Based on the reasoning outlined below, we selected high angle machined grating for all of the instruments.

Replica Gratings: The advantage of a replica echelle grating is that the groove shape is suitable for work at wavelengths as short as the ultraviolet. Hence the groove shape is close to perfect in the IR. From the NOAO spectrograph Phoenix, we have experience with replica gratings on aluminum substrates. These gratings are durable and can be cooled to LHe temperatures without debonding of the replication material (epoxy) from the substrate. Replica gratings are readily obtainable at relatively modest cost.

The major disadvantage of replica gratings is that the groove spacing is very fine for use in the IR. The master rulings are made by diamond burnishing a thick gold coating. Due to the displacement of gold, the coarsest groove spacing possible with this technique is ~ 20 lines/mm. The free spectral range (FSR) from a 20 line/mm R2 echelle is ~ 110 cm-1. At R = 120,000, a pair of buttable 2 K x 2 K arrays (i.e., 4 K dispersive elements) covers only 77 cm-1 at 4500 cm-1 (the middle of the K-band). In other words, a larger detector format would be required to detect the full extent of the grating's FSR. At longer wavelengths, the FSR becomes even more problematic relative to the array. Clearly, the FSR supplied by available groove spacings on replica gratings is too big for the large format detector arrays envisioned. We therefore concluded that the replica grating option was not suitable for the main dispersive element due to this mismatch between FSR and the desire for full spectral coverage.

Immersion Gratings: In principle, immersion gratings can be manufactured with essentially any groove spacing. Thus the FSR can be appropriately matched to the detector size. The use of a high index material for the immersion material also reduces the beam size by a factor equal to the index of refraction for the material. This reduction of the beam size in turn scales down the size of the entire instrument and makes immersion gratings a very interesting option to explore.

However, we believe that there are substantial practical difficulties to implementing immersion gratings. We know of no high-resolution cryogenic instrument that is currently using immersion gratings, although we are aware of at least two groups exploring the development of immersed Silicon gratings (the Pennsylvania State University and the University of Texas).

Four specific problems with immersion grating technology can be identified: groove shape, absorption in the immersion materials over the wavelength range of interest, availability of suitably large crystals of the required material, and cryogenic performance. The groove shape will no doubt be improved as the technology becomes more advanced, and we were not concerned with this difficulty. Our major objection is based on the utilized materials. In the 25-micron region, silicon cannot be used because of absorption features. Germanium is also starting to become less efficient for wavelengths longward of 15 microns, dropping from 96% internal transmittance at 10 microns to about 5% at 17 microns. However, future developments may mitigate some of these issues and make such gratings an attractive alternative for the design of at least the shorter wavelength instrument, NIrES.

Machined Gratings: Diamond turning (rather than ruling) techniques can currently produce gratings suitable for use in the mid-IR. Hyperfine, a company that produces such large format gratings, is confident that the groove shape and spacing are currently good enough for use in the 2 micron near-IR.

We therefore considered two options: standard R2 or R4 gratings with groove spacings appropriate for the desired free spectral range; and high blaze angle R10, coarsely ruled gratings also providing the desired FSR. An interesting aspect of diamond turning technology is that it favors the production of higher blaze angle gratings because, unlike burnishing, the material is cut away rather than relocated within the groove. The use of high blaze angle is also advantageous because the collimated beam size for the spectrograph varies as the tangent of the blaze angle. The change from an R2 to an R10 grating results in a factor of 5 reduction in the beam size for a given spectral resolving power. We therefore concluded that a high blaze angle (R10) machined grating is currently the best choice as the dispersive element for a high- resolution mid-IR spectrograph. We point out that other groups have also independently reached this same conclusion (Ames, U.T.) and have fabricated and employed gratings similar to the one proposed here.

Cross Dispersion: In a machined grating, the groove spacing can be selected to adjust the free spectral range to fill the selected detector in the dispersion direction. A cross disperser is then used to map the entire spectrum onto a two-dimensional array by separating out the overlapping orders of the echelle. Although the optical solution for a cross disperser is often a prism, we have selected cross dispersing gratings to avoid the use of large transmissive elements which could provide cryogenic difficulties. The cross dispersion gratings turn out to have groove spacings easily available in replica gratings. Because the cross disperser is not used at a high blaze angle, the required grating sizes for such replica gratings are also currently available (e.g., from Richardson Grating Labs).

Impact on Flexure: The goal of mapping the entire spectrum onto the detector at one time has advantages in the mechanical design. There is no need to adjust the angle of either the echelle or cross dispersing grating. This eliminates any possible motion in the grating resulting from drift in the mechanism and allows the firmest possible mounting for the grating. In turn, this allows the construction of a mechanically rigid spectrograph that has been designed to minimize flexure, an important aspect that helps reduce the complexity and thermal mass of these large instruments. THE NIRES SPECTROGRAPH -- A 1-5 MICRON ECHELLE SPECTROGRAPH

The near infra-red echelle spectrograph (NIrES) is a high-resolution spectrograph fed with diffraction-limited images at the Cassegrain location of the GSMT. It is fed with an f/18 focal ratio and has the following parameters and components:

The instrument includes foreoptics that allow the light to be split in wavelength with a dichroic for illumination of the spectrograph (near-IR component) and for illumination of wavefront sensors and acquisition TVs (optical component). An Offner relay (1:1) is included to provide a cold pupil for a pupil stop and the implementation of order separation filters. The slit face will be tilted in order to allow acquisition and guiding with IR cameras that view the slit and run at video rates.

Optical Model: The optical layout for the NIrES instrument is shown in Figure 1. The 8-element camera is shown in better detail in Figure 2. The camera includes elements of CaF2, BaF2, and ZnSe, with a set of elements that move for focusing between the two spectral regimes (1.6 - 2.4 and 3.5 - 5.0 microns).

Figure 1   Optical layout of the NIrES spectrograph. Following the light path from the bottom are the
slit, the off-axis collimator, the echelle, the cross disperser, the 8-element camera, and the detector. Figure 2   Detail of the 8-element camera. Optical materials include CaF2, BaF2, and ZnSe. The largest
element is about 150 mm in diameter. Light passes from the right to the left.

Figures 3 and 4 display the diffraction-limited performance of the design at 1.7 and 2.5 microns.

The diffraction-limited spectrograph described here has a smaller collimated beam (40 mm) than many spectrographs now in use at large telescopes that are not diffraction-limited. This is due in part to the selection of an R 5 grating with ~ 9 lines per mm to match the FSR to the 2 K pixels in an Orion InSb detector. The optimum cross disperser is a grating of ~ 110 lines/mm and low blaze angle, but other gratings with similar groove spacings could be used. This combination allows the 1.6 - 2.4 micron region to be imaged onto a single Orion array in a single exposure. This clearly demonstrates that the concept can be implemented. However, a more ambitious design would be to use a mosaic of detectors and to image more orders on the array, perhaps the entire 1.5 - 5 micron region. The inclusion of more orders makes the design more challenging because the camera must function over a larger field of view, and a prism would be needed for cross dispersion in order to allow more than one octave of spectral coverage.

Figure 3   RMS spot sizes for wavelengths near 1.7 microns. The wavelength range is that covered
in one spectral order on the detector. Figure 4   RMS spot sizes for wavelengths near 2.5 microns. The wavelength range is that covered
in one spectral order on the detector.

We note that a filter wheel must be provided at the Lyot stop to block the unwanted orders of the cross disperser. Narrowband filters provided here would also allow the observation of single orders with a longer slit. The strawman design presented here would need changeable cross dispersers to adjust the wavelength range between the 1.6 - 2.4 and 3.5 - 5.0 micron regimes, but it is likely that a cross disperser could be found that would serve all of the wavelength ranges in different orders by using different separation filters.

The spectrograph will need to be cooled to about 60 K to eliminate noise resulting from 5-micron thermal radiation. This temperature can easily be achieved using closed-cycle refrigerators.

Sensitivity: With current epoch detectors, and assuming 10% throughput, it would be possible to map the entire 2-micron region spectrum of a K ~ 18 star in a 1-hour exposure with s/n ~10. With the next generation detectors, this should be improved by at least one magnitude. The magnitude scale is, of course, biased to the optical because the brightest IR sources are K ~ -4. Thus for reddened objects or objects with low effective temperature, a spectrograph of this type would far exceed the sensitivity of an optical instrument like HIRES (high resolution echelle spectrograph). THE MIHDAS SPECTROGRAPH -- A 16-20 MICRON ECHELLE SPECTROGRAPH

The MIHDAS (mid-IR high dispersion AO-fed spectrograph) is a high-resolution spectrograph fed with diffraction-limited images at the Cassegrain location of the GSMT. It is fed with an f/18 focal ratio and has the following parameters and components:

The instrument includes foreoptics (similar to the NIrES instrument) that allow the light to be split in wavelength with a dichroic for illumination of the spectrograph (mid-IR component) and for illumination of wavefront sensors and acquisition TVs (optical or near-IR component). An Offner relay (1:1) is included to provide a cold pupil for a pupil stop and the implementation of order separation filters. The slit face will be tilted in order to allow acquisition and guiding with IR cameras that view the slit and run at video rates.

Figure 5   Optical layout for the MIHDAS concept. Top and side views are shown.
The R10 echelle is the rectangle located at the upper left in the top view. The dimensions for the
echelle are 150 x 1500 mm. The overall length of the instrument is about 4 m! The camera is a four-
mirror system.

In the mid-IR, the diffraction-limited image is large (about 0.25 arcseconds), resulting in a large collimated beam at the echelle grating. At 20 microns, the design requires a collimated beam diameter of 0.15 m in order to achieve the desired resolving power with an R10 echelle. The collimator has focal length of 2.7 m! The R10 echelle is 0.15 m wide by 1.5 m long. Hyperfine has already machined echelles of similar size so the grating could be obtained with today's technology. The camera elements must be larger than the collimated beam diameter to accommodate the field imaged on to the detector. We expect that the instrument as described here would require a cryogenic enclosure of about 4 x 2 x 2 m! Paraxial calculations show that by using four butted 1024 x 1024 arrays covering an area of 55 x 55 mm, the entire 16.5 - 20 micron region can be mapped onto the detector. Order separation is at minimum 11 pixels, allowing a 1 arcsecond slit length for adequate background subtraction with the 0.25 arcsecond image.

Figure 5 shows the schematic layout of one possible design for this instrument in which the camera is a set of four off-axis mirrors, each having high order aspheric surfaces. Unfortunately, the optical design has not yet been fully optimized for good images and requires further effort.

This instrument presents two challenging issues. The first is the camera design. We believe this is soluble with the four mirrors but, as stated above, our designs did not fully converge by the time of this report. The second challenging aspect is the cooling. A Si:As detector sees radiation as long as ~ 28 microns. This will require that the grating be cooled to ~ 10 K. Standard closed-cycle refrigerator technology will not reach this temperature, so liquid helium is required. Instrumentation for particle physics has cooled instrumentation of the envisaged size and mass with a liquid helium refrigeration plant located next to the chamber, so the technology exists to implement such an instrument as MIHDAS. However, running such a plant at the telescope, especially on a remote mountaintop in Chile, may prove to be somewhat difficult. Further thermal engineering could be used to explore approaches in which only critical elements are cooled to 10 K, while noncritical elements of the instrument might be left as warm as 20 K. Such approaches have been used successfully in existing instruments, such as the T-ReCS instrument built by the University of Florida for the Gemini telescopes.

March 2002