3 Spectral observations and data reduction

3.1 SAO 6m observations

The first observations of a small subsample of HSS candidates were conducted in January 1995 with the SAO 6m telescope. The telescope was equipped with the 1024-channel photon counter detector (IPCS) mounted on the medium resolution spectrograph SP-124 in the Nasmyth-1 focus (Drabek et al. 1986). A 2$\hbox{$^{\prime\prime}$}$ diameter circular aperture was centered on the brightest part of the target galaxy or on the star-like candidate, and a similar aperture was placed 20$\hbox{$^{\prime\prime}$}$ off along the direction zenith-nadir to accumulate simultaneously the night sky spectrum. The latter was subtracted during the data reduction from the target spectrum. This set-up resulted in a spectral coverage of $\lambda$ 3700-5500 Å with a scale 1.8 Å/channel along the dispersion and a spectral resolution (FWHM) of about 10 Å. The exposure time varied between 3 and 5 min depending on the brightness. The signal-to-noise ratios of these spectra are $\approx$5 in the continuum at $\lambda$ 5000 Å. The seeing was about 2$\hbox{$^{\prime\prime}$}$.

During the night we have observed spectrophotometric standards from the KPNO list (Massey et al. 1988) which were used to correct the spectra of the program galaxies for the system spectral response. An Ar-Ne-He lamp was used for the wavelength calibration. A flat-field was accumulated every night before and after the observations and all spectra are corrected for the flat field after azimuth and modulation correction of the original 1024-channel arrays.

3.2 Calar Alto 2.2m observations

Further follow-up spectroscopy was made during nights May 30 - June 4, 1995, using the Boller & Chivens spectrograph attached to the Cassegrain focus of the Calar Alto 2.2m telescope. A 300$\hbox{$^{\prime\prime}$}\times2\hbox{$^{\prime\prime}$}$ long slit was used. The spectra were recorded on a 1024 $\times$ 1024 pixel Tektronix CCD operated in a 2 $\times$ 1 binned mode (binning only along the slit direction), resulting in a spectral resolution of 9 Å and a wavelength coverage $\lambda$ 3700-8100 Å. The exposure times varied between 2.5 and 15 min depending on the galaxy brightness. The observations were complemented by standard star flux measurements, lamp exposures for wavelength calibration, dome flats, bias and dark frames. The seeing was mostly about 1$\hbox{$^{\prime\prime}$}$, rising at times up to 2$\hbox{$^{\prime\prime}$}$.The observations during the second and the last night were made under non-photometric conditions. The flux calibration for the dozen objects observed during the second half of the last night is especially uncertain.

3.3 Data reduction

Data reduction was done at SAO with the MIDAS 94NOV software package. The context SPEC is adapted to the SAO data formats to perform an automatic reduction of the one-dimensional IPCS spectra (Kniazev 1994). Corrections for flat field, azimuthal shift and two-channel modulation are applied to the IPCS spectra. Then, standard corrections for the atmospheric extinction, sky subtraction, and the instrumental response curve are done. All the IPCS spectra were taken mainly during rather poor, non-photometric weather conditions and, therefore, are presented in a relative count scale.

For the Calar Alto two-dimensional CCD data, the reduction included bias-dark correction, cosmic-ray removal and flat-field correction. 1-D spectra are extracted by adding 3-4 consecutive CCD rows centered at the object intensity peak. After sky subtraction they are converted to linear scale and corrections for atmospheric extinction and flux calibration are applied. For the flux calibration we use the mean response curve obtained from observations of two standard stars. Calar Alto CCD spectra are presented on an absolute flux scale.

The redshifts and line fluxes are measured applying Gaussian fits to the line profiles. The average redshift of each galaxy is derived from the observed wavelengths of all prominent individual emission lines excluding [OII] $\lambda~3727$ Å, which is outside of the region where the wavelength calibration curve is well determined. The typical internal redshift errors are estimated to be less than 0.0001 for the Calar Alto emission-line spectra and less than 0.0003 for the 6 m emission-line spectra. For galaxies with absorption features the redshift errors are within $\sim$0.0005.

The emission line fluxes are derived as the sum of the pixel intensities inside the line region using standard MIDAS program tools. For all but one spectrum, the individual emission line fluxes of the H$\alpha$, [NII]$\lambda\lambda$6548, 6583 Å and [SII] $\lambda\lambda$ 6716, 6731 Å line blends are obtained by summing of pixel intensities over the total blend and then modelling the individual line fluxes using Gaussian fitting.