KPNO IRMOS INFORMATION
Introduction
The Infrared Multi-Object Spectrograph (IRMOS) is an innovative near-infrared instrument employing an array of MEMS micromirrors at the telescope focal plane for target selection. This instrument was a joint project of the Space Telescope Science Institute, NASA James Webb Space Telescope Project and Goddard Space Flight Center, and KPNO. It is presently offered for use by the astronomical community on the Mayall 4-m telescope.
IRMOS uses a Texas Instruments 848 × 600 element digital micromirror device (DMD) as a focal plane mask to synthesize slits within a 3 × 2 arcmin field of view. Wheels containing filters and fixed-orientation gratings provide R ~ 300, 1000, and 3000 spectroscopy in the J, H, and K bands and R ~ 1000 in the Z band. In addition, the instrument can be used for imaging, although because of field distortion and scattering from the DMD elements, we do not recommend IRMOS for quantitative imaging science.
Instrument Description
A line drawing of the (rather complicated) optics path is
shown in Figure 1. The f/15 beam from
the 4-m telescope is focused outside the instrument window. The focal plane is reimaged by three mirrors
(mirror 1, mirror 2, fold 1) onto the DMD at a 3:1 reduction to give a pixel
scale of 0.2 arcsec. The DMD itself is a
848 × 600 array of individually controllable 17 micron square mirrors, each of
which can be switched from a tilt orientation ~ 20º on one side of the normal
to the input optical axis to ~ 20º on the other side of the normal. One of these positions (“on”) directs the
reflected light to the spectrograph collimator (mirror 3), the other (“off”) to
a dark baffled area which absorbs the light.
The collimated light is folded (fold 2) onto the grating. After dispersion, the spectrum is imaged onto
the 1K × 1K
Figure 1. Line drawing of the IRMOS optical components
and light path. IRMOS employs mirrors as
all of the active optical elements.
The DMD mirrors are individually controlled by a Digital Signal Processor through the IRMOS host computer. One may select all of the mirrors to be “on” for imaging or “off” for measuring the instrumental background or program defined slits anywhere in the field. Because of scattering from the edges of the individual mirrors, the fill factor with all mirrors “on” is not 100%, and there is some signal from bright sources in the field even with the mirrors “off”. The “on/off” contrast ratio is ~ 400. The DMD is also not capable of withstanding cooling to the ~ 100 K bench temperature, so it is maintained at a temperature of 240 K, where it is known to operate without any damage. This does, however, generate thermal emission which limits the performance of IRMOS in the K band.
Filters and Gratings
The various imaging and spectroscopic combinations are achieved with a set of filters and fixed-orientation gratings on rotary mechanisms. IRMOS provides filters for Z, J, H, and K bands as well as half-band filters for J, H, and K to support the R ~ 3000 spectroscopy (otherwise the spectra are too long to fit on the detector). Each spectroscopic configuration uses a separate grating to eliminate the need for a more complex grating mechanism to provide continuous motion of the grating angle. The tables below list the IRMOS filters and gratings.
IRMOS Filters
|
FILTER |
50% Blue (nm) |
50% Red (nm) |
|
Z |
847 |
1142 |
|
J |
1131 |
1339 |
|
J1 |
1125 |
1258 |
|
J2 |
1216 |
1348 |
|
H |
1431 |
1803 |
|
H1 |
1457 |
1609 |
|
H2 |
1589 |
1823 |
|
K |
1909 |
2456 |
|
K1 |
1909 |
2200 |
|
|
2119 |
2460 |
The resolutions given in the grating table are based on a 3-pixel (0.6 arcsec) slit. As with all focal plane MOS instruments, the spectral coverage can depend on the location of the slit within the field. The R ~ 3000 gratings produce spectra which will fall off the detector for slit locations which are more than approximately 0.5 arcmin from the center of the field, so the useful acquisition field for full coverage of the bandpass is approximately 3 × 1 arcmin. The numbers in black represent those measured on the telescope (see Spectral Resolution section); those in red are calculated and have not yet been verified.
IRMOS Grating Configurations
|
CONFIGURATION |
DISPERSION (Å/pix) |
RESOLUTION (λ/dλ) |
|
J 300 |
|
188 |
|
H 300 |
|
246 |
|
K 300 |
|
327 |
|
Z 1000 |
|
1063 |
|
J 1000 |
3.854 |
1260 |
|
H 1000 |
|
1050 |
|
K 1000 |
9.073 |
880 |
|
J1 3000 |
|
3140 |
|
J2 3000 |
|
3450 |
|
H1 3000 |
|
2630 |
|
H2 3000 |
2.121 |
2670 |
|
K1 3000 |
2.846 |
2430 |
|
|
|
3090 |
Calibration
A calibration unit containing continuum and arc lamps is built onto the top of IRMOS. This permits one to obtain flatfield calibrations using the same slit configuration and at the same telescope orientation as used for science observations. In practice, there has been little experience with the arc lamps, and the night sky OH lines have generally been used for wavelength calibration.
Host Computer
IRMOS is run from a dedicated Windows PC providing graphical user interfaces for controlling all aspects of IRMOS operation, from initializing the detector, setting filter/grating/DMD configurations, taking data and displaying the images. Real-time images of the field can be obtained and the user can select objects for spectroscopy with the mouse. The dimensions of the slit (in units of DMD pixels) are fully definable, and tools to edit the resulting slit mask and to image it (normally inverted) on the target field are provided. Once the slits are set up, the DMD mirrors are reset to normal mode to reflect only the light from the slit mirrors into the spectrograph.
At present, the telescope information is not encoded into the image header and there is no ability to generate telescope offset sequences. A general scripting capability to automate IRMOS observations is still under development.
Useful Facts
|
Wavelength Coverage |
850 – 2500 nm |
|
Spectral Resolutions |
300, 1000, 3000 |
|
Pixel Scale |
0.2 arcsec (4-m telescope, f/15) |
|
DMD |
848 × 600 element (170 × 120 arcsec) |
|
Detector |
Hawaii-1, 1K × 1K HgCdTe |
|
Detector Gain |
~ 4.8 e/ADU |
|
Read Noise |
~ 2.6 ADU (12.5 e) |
|
Full Well |
~ 25000 ADU |
Imaging Performance
As noted above, the scattered light from the DMD mirrors and the field distortion, as well as the limited field of view, do not make IRMOS well-suited for quantitative imaging science. Nonetheless, it does function quite well as an imager, as this is key to the process of real-time definition of slits on the target field. IRMOS can also be pressed into service, if necessary, as an imager for Target of Opportunity observations.
The table below provides some observed signal levels and backgrounds from runs in March and November 2006. The diffuse background seen with the DMD “off” is scattered light at short wavelengths and thermal emission from the DMD in the K band. The installation of a baffle prior to the November run reduced the short wavelength scattered light somewhat. The K band backgrounds were somewhat higher in November, possibly due to warmer ambient temperatures as well as the DMD emission. We plan to experiment with lowering the DMD temperature to 230 K, which should reduce the thermal emission significantly. The higher background for imaging in the K filter also results from the longer red cutoff (2460 nm) compared to that of the photometric K filter (2400 nm).
IRMOS Imaging Signal and Background Levels (ADU-s-1)
|
FILTER |
SIGNAL (10 mag) |
BACKGROUND (DMD ON) |
BACKGROUND (DMD OFF) |
BACKGROUND (mag-asec-2) |
|
J |
3.86 × 104 |
11 - 24 |
2.5 – 3 |
14.6 – 15.7 |
|
H |
4.90 × 104 |
30 – 45 |
2.5 – 3 |
14.2 – 14.6 |
|
H2 |
|
30 |
|
|
|
K1 |
2.93 × 104 |
45 – 70 |
5.2 |
13.1 – 13.7 |
|
K |
4.49 × 104 |
140 - 400 |
31 |
12.0 – 13.0 |
Spectroscopic Performance
The spectroscopic performance has not been tested in all configurations, but the following results obtained during the November 2006 observing run should be a useful guide to those contemplating IRMOS observations. Observations of standard stars and the planetary nebula NGC 7027 in the J, H2, K1, and K bands were carried out to determine the sensitivity and spectral resolution performance as well as to experiment with different observing protocols and calibration procedures for efficient data reduction.
Observing Strategy
Traditionally, the best strategy for IR spectroscopy (both longslit and MOS) is to use a slit of sufficient length to allow beamswitching the target between two positions in the slit. Subtraction of temporally adjacent images removes the bias structure and dark current in the detector and most of the sky/telescope emission, depending on the stability. For extended targets, one often is required to offset the telescope to blank sky to obtain the sky information, incurring a time overhead of at least 33% (one can use an on,off,on,on,off,on,on… sequence to maintain temporally adjacent sky frame). With IRMOS, two additional strategies for observing extended targets are possible. One can obtain “off” spectra by turning all of the DMD mirrors “off” while continuing to guide on the target. In this case, the detector and internal background can be subtracted, but sky emission is not; one must reduce the data as with optical spectroscopy, by flatfielding and subtracting the sky emission spectrum from that of the target+sky. A second technique for moderately extended objects is to define a second (or more) set of slits offset from the first and beamswitch between them. As with traditional on-slit beamswitching, this technique permits one to subtract the sky emission as well as the detector and internal background.
These two cases are illustrated below. The first shows the setup of two slits on NGC 7027, the “on” and “off” images, and the difference. The configuration was the K1 3000. Note that the difference spectrum contains the sky lines, which are superposed on the NGC 7027 spectrum and must be removed by separately extracting the sky spectrum or by extracting the object spectrum in background subtraction mode with the background region carefully defined. There are two defects in the DMD presently installed in IRMOS; one of them can be seen in the acquisition image below. These will create their own “spectra”, which show up approximately 20% of the way in from the left and right edges of the array.
Figure 2. Illustration of the “DMD on-off” method of
observing an extended target. The left
panel shows the setup of two slits on NGC 7027.
The slits are sufficiently long to extend well off the target to obtain
sky spectra. The bright spot in the
lower right is a defect in the DMD; another is off the figure to the bottom
left. The next two panels show the
spectrum of NGC 7027 with the DMD mirrors “on” and “off”, respectively, and the
right panel is the difference image. The
bright background is dispersed continuum from the warm (240 K) DMD. The K1 3000 configuration covers the region
1930 nm (top) to 2230 nm (bottom).
The second strategy involved setting up three identical sets of the two slits used for NGC 7027 and beamswitching the telescope between these positions. Figure 3 shows the same sequence as in Figure 2, except that the middle panels show target observations on the center and right set of slits. The most obvious advantage of this strategy is that one is always observing the source. In addition, by subtracting the majority of the sky background prior to flatfielding, the residual sky lines are eliminated much more efficiently.
Figure 3. Illustration of beamswitching using three
sets of slits programmed into the DMD (left panel). Succeeding panels show observations on two
of the three sets of slits and the difference, which virtually eliminates the
OH sky lines.
The strategy of using the DMD to obtain “off” images does not seem to provide any advantages, even in the case of a target which is sufficiently extended that one cannot set up another set of slits within the IRMOS field. One is always left with the problem of subtracting the strong sky spectra from the target, which is critically dependent on achieving superb flatfielding. In the case of large targets, it is probably best to simply offset the telescope to a sky position and obtain true sky frames. For modestly extended objects, the multiple slit technique illustrated for NGC 7027 appears to work well and is operationally trivial. Once one has defined the slits on the targets, it is very easy to define additional sets in the DMD mask configuration file.
Signal Levels and Performance
We have not yet measured signal levels in all configurations under photometric conditions, but the results presented here for four configurations should provide a rough guide for estimating the performance. The signal levels were measured using a photometric standard and a 3-pixel (0.6 arcsec) slit, which will probably be representative of most IRMOS applications. Unfortunately, the conditions were not uniformly photometric. Furthermore, the noise which is used in the performance estimates was a function of the observation technique, as noted above. The K 1000 spectra were obtained only on the night where the “DMD on-off” technique was used, and this was determined to be noisier due to the incomplete subtraction of the sky background. The noise was determined by extracting an “off-source” spectrum using the same extraction aperture as for the standard on the long NGC 7027 exposures and scaling to 1 hour. Because of the partially photometric conditions, these estimates are quite likely conservative.
The signal levels are given in ADU-s-1 for a 0.0 mag star.
Figure 4. Spectra of an A0 standard star scaled to 0.0 mag in ADU-s-1. Configurations were J 1000 (left) and K1 3000
(right).
IRMOS Signal/Performance
|
CONFIGURATION |
SIGNAL (0.0 mag) |
BGND (ADU/s) |
S/N = 10 1 hour |
|
J 1000 |
300000 |
~ 1 |
15.2 |
|
K 1000 |
380000 |
0.8 - 125 |
13 – 14.4 |
|
H2 3000 |
90000 |
0.6 – 1.2 |
14.9 |
|
K1 3000 |
110000 |
1.5 - 8 |
14.9 – 15.4 |
Spectral Resolution
The spectral resolution was determined from both the NGC 7027 and OH airglow spectra; representative spectra, all taken with a 3-pixel (0.6 arcsec) slit, are shown below. In general, the spectral linewidth FWHM was very close to 3 pixels, yielding the spectral resolution given in the earlier table. Both the OH and NGC 7027 spectral lines show a slight extended wing on the blue side, which can be seen in the spectra presented in Figure 5. The cause of this has not been determined; although this does not appreciably affect the line width, it could affect integrated line flux measurements by at the 5% level. We will be investigating this effect as time permits.
Figure 5. Spectra of NGC 7027 at R ~ 3000 in the H2
(left) and K1 (right) filters. Spectral
features at the .001 level relative to the H I 4-7 line (Br γ) can be
easily detected. Most of the “noise” on
the spectra are residual OH sky lines (at roughly .004 Br γ) which were
incompletely removed by sky subtraction because of the non-photometric
conditions.
Additional Information
For those who are interested, files of the plots presented here (and some additional ones) are available for more detailed inspection or downloading.
NGC 7027 DMD on-off difference
NGC 7027 center – right difference
The manual for setting up the IRMOS computer and monitoring the temperature during the cooldown may be downloaded from the link below:
Updated 7 August 2007
NOAO Contact: joyce@noao.edu