2 Observations and data reduction

The objects observed in the present program were selected from General Catalogue of Variable Stars-GCVS and from New Catalogue of Suspected Variable Stars-NSV (Kukarkin et al. 1969 and supplements; Kholopov et al. 1985, 1987; Kukarkin et al. 1982). A total of 616 variables of the types L, L:, I, I:, IS and IS: was observed photometrically. The results of this survey are published in a separate paper (Cieslinski et al. 1997a, hereafter Paper I). If a star presented a colour index B-V<1.4 or a catalogued photometric variation $\Delta m\gt 1.5$ mag, it was included in our spectroscopic program. Possible reddening effects were also taken into account by correcting the B-V colour with a standard correction dependent on galactic latitude.

The selected objects were observed in two missions (July 22-24/1985 and June 22-July 04/1986) with the Boller & Chivens Spectrograph at the 1 meter Yale telescope of CTIO. The detector used in the 1985 mission was a GEC CCD, which provided a wavelength range from 4500 to 7000 Å, with 5 Å resolution. In 1986 the detector was the 2D-Frutti, which gave a spectral coverage from 3900 to 6800 Å, with 5 Å resolution as well. Two of the stars, KQ Mon and V499 Ori, were observed in November/1986, with the same instrumental configuration of the June-July/1986 mission. The reduction was carried out with the TVRED package at CTIO.

Additional spectra were collected at Laboratório Nacional de Astrofísica (LNA/CNPq), in southeast Brazil. In this case we used a Boller & Chivens Cassegrain spectrograph at the 1.6-m telescope. The detector was a UV coated GEC P88230 CCD, with $770\times1152\;{\rm pixels}$ and 22.5 $\mu$m/pixel. Two gratings with 300 and 900 lines/mm were used. The wavelength coverage is about 4650 Å and 1350 Å, with 10 Å and 3 Å of resolution, respectively for the 300 and 900 lines/mm grating. The reduction of these spectra was done using the IRAF software installed at the SUN workstations of the Astrophysics Division of INPE.

Spectrophotometric standard stars (Oke 1974; Stone 1977; Stone & Baldwin 1983; Baldwin & Stone 1984; Taylor 1984) were observed each night to flux calibrate the spectra. Our spectra shown in Fig. 1 to Fig. 5 are flux calibrated. However some caution should be taken with the accuracy of these calibrations. Many of the nights were non photometric and unknown slit losses, specialy in the blue, make our absolute calibrations never better than about 30$\%$. The spectra obtained with the 2D-Frutti display a significant blue deficiency, quite evident in these figures. Such distortion is related to the difficulty in flux calibrating 2D-Frutti data. On the other hand, our main interest is to classify the objects and not to derive absolute fluxes values.

The variable stars observed spectroscopically (a total of 169) are listed in Table 1. The name of the object is given in Col. 1; Col. 2 gives the Julian Date of the beginning of the observation and Col. 3 the exposure time. A tentative spectral type, consigned by comparing visually each observed spectrum with those presented in libraries of stellar spectra (see, e.g., Jacoby et al. 1984; Pickles 1985; Kirkpatrick et al. 1991; Silva & Cornell 1992; Torres-Dodgen & Weaver 1993; Danks & Dennefeld 1994; Fluks et al. 1994), is indicated in Col. 4. The mean B-V colour is shown in Col. 5 while Col. 6 gives the site of the observation.