We acquired medium resolution spectra for stars with spectral types earlier than 
M0 and low resolution spectra for late M type stars. Only a few bright stars having well
documented spectral classification available in the literature were not observed.
Because of the relatively short wavelength range in the medium resolution mode
(
500 Å) computerized algorithms involving theoretical reference
spectra are not suitable (Cayrel et al. 1991). Instead we extensively
used the classification methods presented in Jaschek & Jaschek (1987)
and Turnsheck et al. (1985). Our spectral type determinations result from a visual
comparison of our spectra with those of MK standards. For this purpose we used the
library of Jaschek (1992); Turnsheck et al. (1985) and Jacoby et
al. (1984). In addition we also observed few MK standard stars in the range
with the same instrumentation, wavelength range, and resolution as
for our candidate stars.
Various classification criteria were considered depending on whether a blue medium
resolution or a low resolution spectrum was available and depending on whether the star
had an early or a late spectral type. For stars of spectral type earlier than F, the
classification depends on the strength of the 
H&K lines, which become much
stronger than the hydrogen lines of the Balmer series. For spectral types later than G5
the 
line (
4227 Å) becomes highly sensitive to temperature.
Metallic lines increase both in number and in intensity and are the main criteria of
classification toward later spectral types. One of the major difficulties concerning
the classification of our candidates is that for these active stars, a fraction of the
energy is re-emitted in the Balmer lines. Therefore, all spectral determinations
involving H
(a major indicator in our wavelength range) may undergo a shift of
several tenths of spectral subtype. Whenever possible, the H
line has not been
taken into account in the spectral type determination. We have also checked for luminosity
effects using 
lines (
4077 Å and 
4215 Å) but no
significant enhancement has been detected implying that most of our stars are main
sequence stars. Although this is consistent with results from former Einstein stellar
surveys (see e.g. Rosner et al. 1985), we cannot exclude that a
small fraction of our candidates have erroneous luminosity classes. For most M
stars only low resolution spectra were available and the main criteria of
classification were the strength of the CaOH and TiO molecular bands.
Comparing our own spectral type determinations with those quoted in the literature for stars observed in various RGPS test areas or X-ray selected samples, we estimate that the uncertainty on the spectral classification should not exceed two spectral subtypes.
Visual magnitudes were either extracted from the SIMBAD database or from the Guide Star
Catalogue (Lasker et al. 1990). We also list the B-V colour index
in Tables 4 (click here) and 5 (click here) when available from the literature. For all stars
having a spectral type determination we corrected the V magnitudes taken from the
GSC for colour effects according to relation (1) of Russell et al.
(1990), assuming a B-V colour index corresponding to the spectral type
and neglecting interstellar reddening. The magnitudes of GSC stars without
spectral types were left unchanged. In order to estimate the photometric quality
of the corrected GSC magnitudes we compared these GSC based determinations with V
magnitudes extracted from SIMBAD header and associated literature. SIMBAD values
generally arise from photometric measurements and have therefore much smaller
associated errors than the photographic determinations used in the GSC. Based on 33
stars found in the course of RGPS optical identifications in various test fields
and having both GSC and photometric SIMBAD measurements we conclude that ignoring a
few pathological cases for which the difference in magnitudes is larger than 0.6
mag, the difference 
had a mean value of
0.047 and a rms of 0.18 down to m = 13.5. We thus adopted a 1
error of
0.2 mag as statistically representative of the uncertainties affecting the colour
corrected GSC V magnitudes. For stars without spectral types we assumed a
photometric error of 
0.3 mag (Russel et al. 1990). For the
few stars with no SIMBAD nor GSC identification (generally Me stars fainter than 
) we derived visual magnitudes by differential photometry with nearby
GSC stars using V CCD images. The corresponding photometric errors are dominated
by those of the comparison GSC stars and are therefore of the order of 0.3 mag.
Table 1: Characteristics of the CCD used for
spectroscopy and imagery
As for photometric data, the coordinates of the proposed optical counterparts were
preferentially extracted from the SIMBAD database which most of the time lists high
quality determinations (e.g. PPM). For the candidates without accurate SIMBAD
positions we used the GSC coordinates. The coordinates of the remaining counterparts
fainter than the GSC threshold were measured on CCD images with respect to nearby GSC
stars. The maximum positional error for the overall identified set is therefore
comparable to that internal to the GSC which is typically less than 2
\
(Russel et al. 1990; Egret et al. 1992). However, some
high velocity objects may have a somewhat larger positional error since neither
SIMBAD nor GSC coordinates are corrected for proper motion. In most cases, the
error on the optical position is expected to be an order of magnitude smaller than
the X-ray 90% confidence radius which is in the range of 17
to 58
(see Tables 2 (click here) and 3 (click here)).
H&K fluxes
The 
H&K chromospheric emission lines are a very sensitive indicator of
stellar activity. Studies based on Einstein X-ray data revealed that the
luminosity of these lines correlate well with X-ray luminosity (e.g. Maggio
et al. 1987). We extensively used this correlation in order to quantify the
likelihood of our stellar X-ray identification (Guillout 1996;
Motch et al. 1997). 
H&K re-emission fluxes were
measured on the blue flux calibrated medium resolution spectra using MIDAS
routines. The photospheric level was estimated by fitting Gaussian profiles to the
broad wings of the absorption profiles. This allows a reasonably accurate
measurement of the residual re-emission fluxes even in the cases of low
chromospheric to photospheric contrast.
Although the blue medium resolution spectra are in principle flux calibrated, the
H&K fluxes extracted from these spectra have to be corrected for the
unavoidable spectrophotometric errors resulting from unnoticed clouds, changing diffuse
absorption in the UV or from erratic light losses in the spectrograph slit due to
variable seeing conditions. For each candidate star we computed the 
4000 Å\
to 
(V band) mean flux ratio corresponding to its spectral
type. This flux ratio was estimated using digitized spectra extracted from the
spectrophotometric library of Jacoby et al. (1984). The true mean flux
at 
4000 Å was then computed using as
reference flux level the V magnitude. Finally, the
H&K emission lines fluxes were multiplied
by the ratio of the expected 
flux to that measured from our ``flux calibrated" medium
resolution spectra. In all cases we assumed negligible
interstellar absorption for these nearby sources (see next subsection).
In the survey Cygnus region, most of the detected X-ray stars are closer than 300pc
(Motch et al. 1997). At this distance, A
is always smaller
than 0.5 over the whole investigated field (Neckel & Klare 1980)
and we decided not to apply additional photometric correction for interstellar
absorption. Absolute magnitudes were derived from spectral type using the
calibration of Schmidt-Kaler (1982). Taking into account the
uncertainties on the V magnitudes and absolute calibrations we estimate that the
photometric distances have an error of the order of 20% yielding an additional
40% error on the X-ray and 
H&K luminosities.