2 Observations and reduction

2.1 The sample

Our sample comprises to a high degree of completeness (90%) all stars in the Bright Star Catalogue fainter than V=5.1 within the spectral domain B0 to A9, in the R.A. range 23-8 hours and south of declination 10$^\circ$ which corresponds to the concept of "southern stars" used in the title of this paper. Occasionally, brighter stars have been included, as well as stars with later spectral types (up to F5) and slightly outside the right ascension domain indicated.

Figure 1 shows the distribution in galactic coordinates of the 803 stars of this sample. Since the great majority lies in the third galactic quadrant which is nearly empty of interstellar dust out to 500pc (Lucke 1978), reddening plays an unimportant role for main sequence A type stars, but has to be accounted for in the case of B type stars (Crawford-calibration and Q-method) and A-F supergiants (intrinsic colours corresponding to spectral types). We used dereddened values (b-y)₀ based on the Stroemgren-data compiled by Hauck & Mermilliod (1980).

  

Figure 1: Distribution of the observed Bright Stars in galactic coordinates

2.2 Observing runs

The measurements in the 3 filter system (g1, g2, y) were obtained at the ESO 50cm telescope on La Silla by two of us (NV, MFA) with the assistance of Messrs. Alberto Blest, Joaquín Perez and Gorki Román. Contributions to this programme were made on 59 nights between September 10, 1974 and June 30, 1976 (Table 1 in the Appendix). The instrument configuration was the same as described in Maitzen & Vogt (1983) and their filter set of system No. 2 was used.

The majority of stars were observed only once (86%) and 11% had 2 measurements. The remaining 3% were measured 3 and 4 times and served to determine the zero-point shifts both in a and g1-y between different observing runs.

The results of our photometry together with pertinent published data are contained in Table 2 in the Appendix.

2.3 Line of normality a₀[(b-y)₀)]

In order to minimize the influence of non-normal stars on the determination of the locus of normal stars in the a vs. (b-y)₀ diagram we eliminated all stars with known peculiar types and emission line stars from our sample, using spectroscopic data from the BS and Michigan catalogues. Additionally, in order to avoid uncontrolled complexity we decided to exclude spectroscopic binaries and visual binaries with separations of less than 15 arcseconds and magnitude differences of less than 3. This way 476 objects remained as sample of normal stars.

A second order regression yielded the normality line:

$$a_0=593.3+0.08496(b-y)_0-7.082^{-5}(b-y)_0^2 [{\rm mmag}].$$ (1)

The standard deviation around this line is 4.74mmag. The sample of normal stars together with the normality line and 3$\sigma$lines above and below it are displayed in Fig. 2.

  

Figure 2: Normal stars and the normality line

  

Figure 3: Histogram of $\Delta a$ values for normal stars

Considering only stars with (b-y)₀ bluer than 0.1 the standard deviation drops to 4.5mmag whence we shall register all $\Delta a$-values with 14mmag and more as peculiar in this range. The scatter around the normality line increases to 5.6mmag for the remaining stars. Since this is not to be attributed to random effects (the stars having the same limiting magnitude as the foregoing ones) it could be explained by intrinsic metallicity differences among the cooler stars of our sample.

Figure 3 shows the histogram of deviations $\Delta a=a-a_0$with a binning of 2mmag. It is interesting to notice that the distribution of the positive $\Delta a$-values declines somewhat less steeply from the maximum than for the negative ones. This could be explained by the existence of marginally peculiar stars which escape detection by both the photometric $3\sigma$ criterion and classification dispersion spectroscopy and form a transition population between peculiar and normal stars. Hints for that can be found e.g. in Maitzen & Vogt (1983).

2.4 The g1-y colour index

Measurements in the 3 filter system g1, g2, y yield not only the index a, but also colour differences one of which is g1-y with the longest baseline in wavelength. It is interesting to ask, especially with such a large sample of objects, both normal and peculiar, whether it shows a good correlation with b-y normally used as reference colour when determining $\Delta a$-values. If so, it might be safely used instead of b-y, especially if the latter should not be available.

Figure 4 shows the diagram g1-y versus b-y with the linear regression:

$$g1-y= 490.5+0.6022 (b-y) [{\rm mmag}].$$ (2)

  

Figure 4: Correlation of the g1-y and the b-y colour indices

The average scatter around this line is 6.51mmag and the correlation coefficient 0.9946. Considering that the errors in the abscissa values b-y are of the same order, the very high degree of correlation becomes obvious. This is not straightforward, since also peculiar stars were included, and some differentiation between the bandpasses of b and g1 could be expected. Moreover, no dereddening procedure was applied. If one looks carefully, one will notice a slight preponderance of points above the regression line in the interval $0.050 \leq b-y < 0.150$, hence among the mid A type stars. This effect which is on the average smaller than 0.010 mag might be due to strong Fe I lines in the passband of g1 and the rather high percentage of metallic line stars in this spectral region.