3 Observations and data reduction

The observations were carried on at the European Southern Observatory at La Silla (Chile) using a 1.5-m telescope equipped with a Boller and Chivens spectrograph and a CCD $2048 \times 2048$ pixels with a readout noise of $\pm 7$ electrons. A grating of 225 gr/mm with a dispersion of 330 Å/mm in the first order was used. The CCD has a square $15\,\mu$m pixel, giving a dispersion of about 5 Å/pixel in the wavelength direction. The useful spectral range is about 4900 $<\,\lambda \,<$ 9200 Å with a FWHM of 10 Å. The spectra were taken through a 5 arcsec slit oriented in the East-West direction. The slit width has been chosen in order to minimize the consequences of atmospheric differential refraction and reduce the loss of light at both ends of the spectrum. This is important since a substantial loss of light may lead to an erroneous spectral characterization of the object. Care was taken also in guarantee that the observations were made as near as possible to the meridian of the asteroid.

The spectra of Themis family asteroids were obtained in five observing runs in January/97, March/97, July/97, December/97, January/98 and March/98. The atmospheric conditions were good to excellent during all the observations. The observational circumstances, from EPHEM program (Tholen 1997), are listed in Tables 1, which shows the distance from the Sun, from the Earth, the solar phase angle, the estimated visual magnitude and the diameter. The given diameters are from IRAS (Tedesco 1997) whenever available, otherwise are estimated through the absolute magnitude and a visual albedo of 0.081 which is the mean albedo of Themis' family objects.

The spectral data reduction was performed using the Image Reduction and Analysis Facility (IRAF) package and taking much care to ensure a proper calibration of the spectra. The bias level of each night was determined through an average of the many bias images taken in the night. This "averaged bias'' was then subtracted from each frame and pixel-to-pixel variations were removed dividing the resulting image by a normalized medium "flat field''. The IRAF apsum package was used to sum the pixel values within a specified aperture and to subtract the background level. Wavelength calibration was performed using a He-Ar lamp, obtained several times during each night, and spectra were corrected from airmass by using the mean extinction curve of La Silla (Tüg 1977). Different solar analogs (Hardorp 1978) were observed in each observational run in order to compute reflectivities. It must be pointed out that in each night of observation were observed at least two solar analogs in order to estimate the quality of the night. The ratios between the spectra of the solar analogs for each night show no substantial variation. The influence of different solar analogs on the resulting spectra has also been checked showing differences less than $1\%/10^{3}$ Å. In Table 1 are given only the solar analogs used to obtain the spectra presented. All asteroid spectra are normalized around 5500 Å by convention, unless otherwise specified.