The Small Main-Belt Asteroid Spectroscopic Survey (SMASS) was initiated in 1990 with the goal of obtaining spectra over the visual wavelength region for a substantial number of small (D < 20 km) main-belt asteroids. A second goal is to make target of opportunity measurements of near-Earth asteroids. Observations are primarily made using the 2.4-m Hiltner telescope, located at the Michigan Dartmouth MIT (MDM) Observatory on the southwest ridge of Kitt Peak in Arizona. We utilize the Mark III spectrograph, equipped with a TI 4849 (398x598 pixels) or Tektronix (1024 x 1024) CCD. Our observations are made with a low resolution grism (150 line/mm, blazed at 7300 A ) to cover the wavelength range from 4000- to 10000-A in a single exposure, with a dispersion of ~10 A/pixel. A Wratten 12 filter (cutoff < 6000 ) is placed over the red dispersion half of the CCD dewar window in order to block the second-order spectral image. This technique has the ability to image the entire spectrum in a single exposure and has many advantages over piecing together the different spectral regions from separate images. In particular, the effects resulting from the asteroid's rotational lightcurve can be neglected when the entire spectral range is imaged simultaneously.
A significant effort is made to ensure proper calibration of the asteroid spectra. Bias images are obtained regularly to monitor the readout level of the CCD, and the flat field characteristics of the CCD are carefully examined. Since our spectral images span the entire sensitivity range of the CCD, we find it difficult to obtain and apply flat field images with a high signal-to-noise at both ends such that the noise introduced is not greater than the intrinsic pixel sensitivity variations. Instead, during the observations, care is taken to place each spectrum along the same CCD columns, minimizing the possible effects of pixel-to-pixel variations in sensitivity. White light flat field images in direct imaging mode show pixel to pixel variations of < 1% over this cosmetically clean region of our CCD. Wavelength calibration is determined from images of Hg, Ar, and Xe lines, obtained from lamps in the spectrograph. Observations are also made of solar-analog stars (usually 16 Cyg B and Hya 64) for calibrating the asteroid's relative reflectance spectrum and of flux-calibration stars to monitor changes in atmospheric extinction. Particular care is taken to avoid calibration problems associated with atmospheric dispersion of asteroid (and star) images. While the typical seeing at MDM gives images with a FWHM of 22 arcseconds, a 4.7 arcsecond-wide slit, oriented in the N-S direction, is regularly used. This slit width results in a spectral resolution of ~50 A. Observations are restricted to within 1.5-hours of crossing the meridian, so that any atmospheric dispersion in the image occurs along the slit direction.
Asteroids are observed down to an apparent V magnitude of 19, using individual exposure times in the range of 15 to 30 minutes. Longer exposures are not used so as to minimize the effects of cosmic ray detection. Whenever possible, each asteroid is observed at least twice as a check for consistency. For comparison and monitoring of possible systematic errors, brighter asteroids with known spectral absorption features are also observed. The telescope is tracked at the asteroid rate of motion, so the asteroid's identity is verified by checking the motion vector with respect to nearby stars.
Spectral data reduction is performed using the Image Reduction and Analysis Facility (IRAF), developed by the National Optical Astronomical Observatories. After bias subtraction, cosmic rays are identified and removed from each image by averaging over the neighboring pixels. The IRAF apall package is used to sum the pixel values within a specified aperture at each point along the dispersion axis, and to subtract the background level, producing a one-dimensional map of pixel intensity (flux) as a function of column position. Wavelength calibrations are performed using spectra from the Hg, Ar, and Xe lamps obtained on the same night. We find there were no significant time-dependent dispersion variations in the spectra obtained from the lamps. Finally, a correction for atmospheric extinction is applied to the data. Although flux calibration errors are minimized due to the relatively large slit used, we did not attempt to obtain absolute flux values from our data. Errors due to uncertainties in the extinction correction are minimized since the standard stars are observed at air masses similar to those of asteroids (airmass difference < 0.1). For faint objects, the spectra are noisy at both the extreme blue and near infrared ends due to the CCD's poor sensitivity in these spectral ranges. After dividing by the spectrum of the solar-type star which is observed on the same night, and normalizing at 5500 A by convention, we obtain the relative reflectance spectra for each asteroid. The systematic errors introduced by the use of two different solar analogs (16 Cyg B and Hya 64) are low.
For more information, see Xu et al. (1995), Icarus 115, 1-35.