Photometry of narrow and intermediate passband constitutes a useful technique for classification of the stars. Although less accurate than the spectroscopic technique, the simplicity of photometric observations, even for faint stars, has led to the design of several photometric systems (Geneva, Vilnius, Walraven, Strömgren, etc.) that measure the spectral characteristics of the stars and allows them to be classified on an HR diagram.
Spectral classification of the stars with the 
system was first
carried out by Strömgren (1966) who defined three regions:
early, intermediate and late according to strength of the absortion in the
Balmer lines, the intermediate group being the stars around the
Balmer maximum, i.e. the spectral range A0-A3 for main-sequence stars.
This division is, in
fact, a classification according to the effective temperature of the star.
The behaviour of the colour indices within the three regions has been described in detail elsewhere (Strömgren 1966) and the classification in these three groups was fully described, and illustrated in Figueras et al. (1991) for main sequence A-type stars. Following the same classification scheme, O- and B-type stars are classified within the early region while F- and G-type stars are classified in the late region.
To cover late A-, F- and G-type stars different calibrations have been
published, and since 
is a good indicator of effective temperature,
we divided the late region into three subregions, according to the
calibrations:
Once the stars has been classified in the regions according to the effective temperature, criteria of luminosity were applied.
The supergiant stars (including luminosity classes I and II) can be isolated
from the
rest of the stars by using a suitable colour-colour diagram. For the early
and intermediate regions
the most suitable diagram is 
and for the late regions the
.
Philip et al. (1976) and Jakobsen (1985) present different limits for
isolating
early supergiants in the 
diagram. These limits are shown in
Fig. 1 (click here),
together with a sample of early-type stars (the sample is described in
the following section). As the separation between supergiants and
non-supergiants is not clearly defined, any criterion are open to discussion.
Jakobsen's relation identifies a large number of
non-supergiant stars
as supergiants, while Philip et al.'s relation is less efficient in detecting
supergiants in the very early region.
We therefore decided to use a new criterion that
combined elements from both relations. We accepted Philip et al.'s
criterion
for stars with 
and modified Jakobsen's relation for stars with
in order to be continuous with Philip et al. The new limit
is also shown in Fig. 1 (click here).
Figure 1: 
diagram for stars with spectral types O-A3.
The solid line represents the criterion given by Philip et al. (1976),
while the long dashed
line represents the criterion used by Jakobsen (1985) to isolate
supergiant stars. The dotted line represents our proposed
modification for stars with 
In the case of the late regions (Fig. 2 (click here)), criteria based on 
(Philip et al. 1976; Olsen 1988) are sufficient to
isolate the supergiant stars up to about F5-type stars (
).
For A4-A9 stars we used 
instead of 
in
order to take into account metallicity and rotation (Guthrie 1987).
For cooler stars the supergiants appear mixed with main sequence stars
on the 
diagram
and the previously used criteria are inefficient in isolating them.
However, the 
diagram (Fig. 3 (click here)) shows that a separation is
possible in terms of an index 
,
which is computed from Crawford's (1975) relation for
luminosity class V, using 
as free parameter. 
isolates those supergiants with 
that cannot be recognized in
the plane 
.
Table 1 (click here) summarizes the criteria adopted for the separation of supergiant stars. The bracket indices were computed according to Crawford & Mandwewala (1976) for each region and each luminosity class.
As with main sequence stars, those stars classified as
supergiants can be classified
into different regions according to their temperature.
As 
is a good indicator of 
(see for example Gray 1992),
the supergiants were divided into groups according to existing calibrations:
Figure 2: 
diagram for the A4-G2 stars. The solid line represents
Crawford's (1975, 1979) standard relation; the dashed line indicates
the limit of supergiants according to
Philip et al.'s (1976) and Olsen's (1988) criteria
Figure 3: 
diagram for A4-G2 stars. Solid line is the standard
relation by Crawford (1975, 1979); dashed line corresponds to
As the giant stars are rather mixed with the main sequence stars in a colour-colour diagram, no attempt was made to isolate the giants. In the present paper "main sequence" includes luminosity classes V, VI and III. Oblak et al. (1976) also recognized the difficulty in isolating luminosity class III from main sequence by studying the mean colours as a function of spectral type and luminosity class.
The full algorithm provides eight photometric regions. The approximate spectral types corresponding to each photometric region are shown in Table 2 (click here).
Table 1: Criteria adopted in considering a star to be supergiant.
The criteria were taken from Philip et al. (1976) for O-A9 stars and from
Olsen (1988) for stars from F0 onwards.
The criteria marked
with an asterisk are those proposed in this paper
Table 2: Approximately spectral types corresponding to the photometric regions.
The regions labeled as main sequence
also include the giants
In order to test the efficiency of the algorithm of classification,
we selected a sample of stars with known spectral type and with complete
photometry. The sources used in building the catalogue were:
photometry for about 40 000 stars,
complete photometry being available for roughly 25 000 of them,
photometry for about 700 main sequence A-type stars,
The main primary sources for spectral types were the Michigan Spectral Survey and the SIMBAD Data Base, operated at CDS in Strasbourg. For stars in HIC (92% of the sample), the agreement between the quoted spectral types and the B-V Johnson index was checked and can consequently be considered an accurate determination of spectral types.
These sources give a total sample of 15 803 stars with complete 
photometry and known spectral type. The number of stars in the sample depending
on spectral type and luminosity class is shown in Table 3 (click here).
Table 4 (click here) gives information about quoted particularities in the spectrum.
Our sample does not include stars with more than one particularity in the
spectrum.
Table 3: Number of stars in the sample according to their spectral type and
luminosity class
Table 4: Number of stars in the sample according to quoted
particularities in the spectrum.
Only stars with a single particularity were considered
Table 5: Percentages of correctly classified "normal" single stars (7 585 stars)
within each spectral type and luminosity class
The classification algorithm was applied to the "normal'' single stars (i.e. non-Am, non-peculiar, without quoted emission lines) with known luminosity class and undoubtful spectrum. The subdwarfs were not considered and the stars quoted as variable in HIC or HM were also rejected. The sample of "normal'' single stars amounted to 8 542 stars (54% of the total sample). The results of the classification were:
index.
Table 6: Classification of non "normal'' stars
The fraction of "normal'' single stars classified correctly within each spectral type and luminosity class is shown in Table 5 (click here). The low detection rate of very early supergiants is evident.
When the algorithm is applied to the remaining stars (Table 6 (click here)) i.e., those with recognized particularities in the spectral type, a large number are classified in agreement with their spectral type, their colour indices being compatible with the standard relations for "normal'' stars. Most of these stars fall in the range of validity of the calibrations and so the algorithm calculates intrinsic colours and absolute magnitudes, although they might be inaccurate. In particular, it is difficult to separate peculiar stars and stars with emission lines from normal stars and for both the calibrations for "normal'' stars are not suitable. Photometric classification confirms the quoted spectrum for about 70% of stars for which the spectrum is doubtful. The photometric classification of 15% of the stars with unknown luminosity class disagrees with their quoted spectral type.