MICS: A New Mid‐Infrared Camera and Spectrometer for Ground‐based Astronomy

The pre‐optics package converts the incident beam from the telescope into a beam with the focal ratio of 10.5, the same as the final one. It consists of a germanium (Ge) entrance window, two Ge lenses, a filter wheel, and a Lyot stop. The incident focal ratios are F/36 for UKIRT, F/18 for the ISAS 1.3 m telescope (The Institute of Space and Astronautical Science, Sagamihara, Japan), and F/45 for the Mount Bigelow 61 inch telescope (Steward Observatory, Arizona). The configuration can be easily changed by exchanging one of the Ge lenses. At UKIRT, the ISAS telescope and the Mount Bigelow telescope, the pixel scales are 0.425, 1.24, and 1.04 arcsec pixel⁻¹, respectively.

The telescope focus is located between the entrance window and the first lens (Lens‐1). The ray enters the Dewar through the window and passes through the first Ge lens, the filter wheel, the Lyot stop, and the second Ge lens (Lens‐2), and it makes an image on the slit position. The Lens‐1 also makes an image of the secondary mirror of the telescope on the Lyot stop to suppress unwanted thermal emission. The beam between Lens‐1 and Lens‐2 is nearly collimated with a diameter of 3.2 mm.

The filter wheel has seven one inch holes. One of them is for a wide N‐band filter for spectroscopy and another is blind for taking a dark image. The other five are used for imaging observations. The filters in use at present are listed in Table 2. Five of them ("9 micron," "10 micron," "11 micron," "12 micron," and "13 micron") are the N‐band standard "silicate" filters manufactured by Optical Coating Laboratory, Inc. All filters are tilted about 2° against the optical axis in order to avoid ghost images by surface reflection of the filters. The filter wheel is driven by a stepper motor mounted on the outside of the Dewar.

The dispersing element of the spectroscopic optics is a concave grating made of aluminum alloy. The surface is coated with gold. The focal ratio of the reflected light is the same as that of the incident light from the pre‐optics.

The concave grating allows us to make the optics small and compact, since collimator and camera optics are not necessary. The grating groove separation was varied to minimize astigmatism and distortion. A similar grating was used for the Mid Infrared Spectrometer (MIRS; Roellig et al. 1994) on the Infrared Telescope in Space (IRTS; Murakami et al. 1996). A detailed description about this type of grating can be found in Harada et al. (1980) and Onaka (1995).

The beam size on the grating is 14.3 mm in diameter and the groove spacing at the center is 10.67 mm⁻¹. The resolving power of the grating is about 100, and the plate scale is 0.047 μm pixel⁻¹ on the detector.

The slit is inserted at the position of the image made by the pre‐optics. The slit width is 150 μm, corresponding to 2 pixels on the detector, and the slit length is 9.6 mm. The slit is driven by an electromagnetic coil. Since the repeatability of the slit position is quite good (<15 μm, corresponding to 0.2 pixel) and there is no other movable component in the spectroscopic optics, the wavelength assignment to each pixel can be regarded as being fixed throughout observations. Therefore, we can minimize the number of wavelength calibrations, allowing efficient observations.

A toroidal mirror is used as an imaging element. Its surface figure is an ellipse whose two foci are located at the center of the detector and the center of the slit position. The plate scale of the reflected light is the same as that of the incident light. This mirror is made of aluminum alloy coated with gold. This is driven by a stepper motor via gears and shafts and they have free gaps. To reduce the gaps and fix the switching mirror position accurately, a pointing mechanism with a spring is employed. When the switching mirror is set into the imaging position, a spring‐loaded stop like a finger pushes the mirror back and removes the free gaps.

The MICS employs a Si:As Block Impurity Band (BIB) array for moderate flux manufactured by Rockwell International. The BIB detector is one of the best detectors in the mid‐infrared (cf. Bharat 1994; Noel 1992). The array format is 128 × 128 with a pixel size of 75 μm. This was the largest format detector for the mid‐infrared when this project started. The characteristics of the detector are summarized in Table 3.

The signals can be read without resetting the signal integration (nondestructive readout), which allows us to apply a multiple sampling method. We usually use subtractive double sampling in which a signal for each pixel is sampled twice, before and immediately after the reset, and the difference between the sampled signals is regarded as the count for the pixel. This method is effective for canceling out fluctuation of reset bias and reducing readout noise.

To read out the detector signals, four clocks and 10 DC biases are required. A block diagram of the electronics system and a schematic drawing of the driver electronics are shown in Figures 3 and 4, respectively. The detector clocks are generated by the Clock Pattern Generator (CPG) board as TTL signals in the VME rack. The CPG board also generates timing clocks for the Analog‐to‐Digital (A/D) converters, the co‐adder circuits, and the secondary chopper controller. The voltages of the detector clocks are shifted to the suitable level of the detector (+3/+7 V) on the detector‐driver board in the analog electronics box which is attached on the Dewar. All DC biases are generated on the same detector‐driver board.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In mid‐infrared observations, a large number of background photons usually falls into each pixel of the detector. The background flux of the imaging with a bandpass filter (Δλ∼1 μm) in the N band is about 2 × 10⁸ e s⁻¹ pixel⁻¹. The full well of the MICS detector is 10⁷ e⁻ pixel⁻¹ and the saturation level of the A/D converters is set as about 7 × 10⁶ e⁻ pixel⁻¹. Therefore, we adopt 40 Hz as the typical frame rate.

The data acquisition system is shown in Figure 4. The detector has four outputs, and the signal from each output is led into each preamplifier. The amplified signals are digitized by four 1 MHz sampling 16‐bit A/D converters.

The preamplifiers and the A/D converters are installed in the analog electronics box with the bias suppliers and the clock level shifters.

The digitized data are led into the digital board, co‐added in the hardware co‐adder circuits, and stored as 32 bit pixel⁻¹ data in the frame memory. The frame memory can be read directly by the control computer. When an integration is completed, stored data in the frame memory are transferred into a storage hard disk of a control computer. The production rate of the data is 66 kbyte per frame and 2.5 Mbyte s⁻¹.

The data are stored in the FITS format, and observational parameters (frame rate, observing mode, detector temperature and so on) are written in the file header.

The whole system is controlled by the computer, which is a Sun‐workstation compatible machine, Force Computer CPU‐5V, and is mounted on a VME rack on the telescope. Observers can control this computer via ethernet from the console of the second computer in the control room.

The MICS control software consists of short programs written in the C language. These programs run in the control computer, and each program makes a simple task as a command line. This method has a great advantage for developers to make and revise the programs. At actual observations, however, command‐line operation may not be convenient. Therefore, a managing software running in the second computer is used for convenient operations. The observers can generate a series of basic commands and send them to the control computer only by inputing the observational parameters, such as the integration time and the chopping frequency, into the management software via a graphical user interface.

Figures 6a, 6b, and 6c show stellar images taken by the MICS to obtain the point spread function (PSF). They are the images of β Peg obtained in the imaging mode with the 9 micron, 10 micron, and 13 micron filters, respectively. Each image is the sum of 39 frames and the integration time of each frame is 25 ms. These data were taken by using the tip‐tilt secondary system of UKIRT.

Zoom In Zoom Out Reset image size

The first Airy rings are visible in all images, which demonstrates that the diffraction‐limited performance has been achieved. The full widths at half‐maximum (FWHM) are estimated to be 1.7, 2.0, and 2.4 pixels (1 pixel = 0425) for 9, 10, and 13 micron images, respectively.

The PSFs of the spectroscopy mode at three wavelengths are shown in Figure 7. These are obtained by slicing a observed spectrum of β Peg along the slit direction. The estimated FWHMs are 2.0, 1.8, and 1.9 pixels at 9, 11, and 13 μm, respectively. These values correspond to the spatial resolution of about 08.

Zoom In Zoom Out Reset image size

We conclude that the MICS has achieved the diffraction‐limited spatial resolution of the telescope diameter in imaging as well as in spectroscopy.

The performance in the spectral resolution was estimated from the line profiles of astronomical objects. Figure 8a shows a spectrum of a massive star‐forming region, W51 IRS2 East, obtained by the MICS (Okamoto et al. 1999). Three fine‐structure lines, [Ar iii] 8.99 μm, [S iv] 10.51 μm, and [Ne ii] 12.81 μm can be seen in the spectrum. Figures 8b–8d display the line profiles. The FWHMs of these lines are less than 2 pixels (∼0.093 μm). Since the intrinsic profiles should be much narrower than 0.093 μm, the instrumental spectral resolution can be derived as 90 at 9 μm and 130 at 13 μm. These values agree well with the designed values.

Zoom In Zoom Out Reset image size

The sensitivity of the MICS when mounted on UKIRT was derived from the observed images and spectrum of β Peg. We adopted the absolute flux of β Peg provided by M. Cohen (private communication).

In the imaging mode, we evaluated the absolute flux scale, based on the measured counts and the stellar absolute flux. The noise was derived from the fluctuation in the count of blank sky region and converted into flux. The noise equivalent flux (NEF) estimated from the fluctuation of the blank sky in imaging mode with the "9 micron" filter is 170 mJy pixel⁻¹ for 1 s integration, or 0.40 Jy arcsec⁻², and thus the NEF for a point source is estimated as 0.30 Jy. This agrees well with the signal‐to‐noise ratio estimated from the fluctuation of the stellar count of each image.

The sensitivity in the spectroscopy mode was derived in the same way as that in the imaging mode. The slit throughput depends on the wavelength, because the slit width (150 μm) is very close to the diffraction limited size. The throughput was estimated as the ratio of the stellar spectrum obtained with the slit to that obtained without the slit, being 61 ± 3% at 9 μm, 58 ± 3% at 11 μm and 56 ± 6% at 13 μm. They can be fitted as (76 - 1.54 × λ)%, where λ is the wavelength in microns.

Using this relation, we estimate that the NEF is 0.49 Jy pixel⁻¹ at 11.0 μm with the bandwidth of 0.1 μm, where the telluric absorptions are minimum in the N band. The sensitivity in the spectroscopy mode for a diffuse source is typically 1.5 Jy arcsec⁻², and that for a point source is 1.4 Jy. The estimated sensitivities are summarized in Table 1.

When a two‐dimensional array is used, many pixels usually observe blank sky. Therefore, the background could be subtracted by the blank sky around the targets in principle. This is so‐called "self‐sky subtraction," and seems effective because the sky is simultaneously measured when the targets are observed.

However, in practice, the background cannot be subtracted by this method accurately in the N band, because of inaccuracy of the flat‐fielding. The background radiation of imaging observations generates 2 × 10⁸ e⁻ s⁻¹ pixel⁻¹, which corresponds to 100 Jy pixel⁻¹ when the MICS is attached on UKIRT. The NEF of imaging observation is 170 mJy pixel⁻¹ (see § 3.5.3). Therefore, the accuracy of the flat‐fielding is required to be better than 0.2%.

The flat‐fielding is made using the blank sky images, which are obtained at different airmass positions. The flat‐field error becomes up to 3% in maximum. This is less accurate than required. Therefore, the self‐sky subtraction is not useful and the measurement of background by the same pixels as that of targets is needed for sensitive observations in the N band. The secondary mirror chopping is one of the most effective methods for it.

To cancel the effect of sky variation, we subtract the sky image obtained immediately before and/or after the target image by using the secondary mirror chopping. The frequency of the chopping must be faster than frequency of the time variation of the sky background pattern. The measurement of the blank sky discussed in the previous section (Fig. 9) suggests that the sky noise is dominant below 0.5 Hz for all wavelengths. Thus, the chopping frequency should be faster than 0.5 Hz.

To demonstrate the effect of the chopping, we took blank sky spectra with various chopping frequencies and derived the noise at each wavelength. The chopping throw was 108. The results are shown in Figure 11. The measured noise almost equals to the shot noise in the range of the frequency above 0.5 Hz, while it is about 3 times larger than the shot noise at the frequency of 0.2 Hz. These results give good agreement with the measurements of the sky noise discussed in the previous section.

Zoom In Zoom Out Reset image size

Because the chopping is performed by wobbling the secondary mirror, different patterns of the background emission from the telescope could appear. This yields a residual pattern after the chopping subtraction is made. The absolute amount and pattern of the residuals depend on the telescope and the chopping throw. In case of UKIRT, the flux of the residuals is about 1 Jy pixel⁻¹, which is almost 6 times larger than the shot noise generated by 1 s integration in the imaging mode. To cancel out these residuals, the nodding of the primary mirror must be carried out. Consequently, this observing process we adopted is so‐called the "chop and nod" technique.

The effectiveness of the nodding depends on its time interval. No residual pattern was detected when an interval of 30 s and a throw of 208 was used. At this time interval, the nodding removes the residuals almost completely (<10 mJy pixel⁻¹). The nodding with intervals of 1, 3, and 5 minutes left the residuals of about 20, 50, and 70 mJy pixel⁻¹, respectively. The interval of the nodding should be shorter than 30 s. This value may vary with telescope, observing condition, and chopping throw.

While the background‐noise limited sensitivity is almost achieved with the chop and nod technique, the observing efficiency is another concern. If the object is compact enough, it is possible to put all images obtained in four beams of the chop and nod on the detector and to avoid the loss of net integration time. This produces four images of the objects, two positive and two negative ones. One can observe objects without the loss in efficiency if they are smaller than 25^'' × 25^'' in imaging mode and 12^'' in spectroscopy mode. If the object is larger, the chopping throw and/or the nodding throw should be larger than the object size. In this case, some of the four images will be put off the detector, and the efficiency will be decreased by a quarter in the worst case.