A reliable spectral type for each programme star is of fundamental importance in order to obtain the accurate information about its extinction characteristics and, consequently, its location in an HR-diagram. In most previous studies only photometric spectral types were available. Such spectral types are only reliable when obtained at maximum brightness, as they can be influenced by anomalous extinction and variability. Furthermore, since photometric spectral types are not always unique, spectral observations are indispensable. However, because of the occurrence of spectral line changes in HAeBes (e.g. Grinin et al. 1994) the spectral type derived from the continuum is certainly welcome.
Two-Colour Diagrams (TCDs) for cluster members can be used to determine the
foreground extinction, to describe a probable common extinction behaviour
and to estimate photometric spectral types by dereddening the observed colours
to intrinsic ones if an extinction law is assumed.
In Thé et al. (1990) the foreground
E(B-V) for NGC 6611 was determined to be at least 0
5.
We have seen from several studies that the extinction law for
each star can be different. The use of a TCD to determine
extinction laws for clusters like NGC 6611, as was done in
Chini & Wargau (1990), is therefore not
reliable. We will use TCDs of the Johnson UBV and the
Walraven ULBV data to estimate the spectral types of our
programme stars and to see if our objects are lying at the foreground.
In Fig. 2 (click here) the TCD based on the UBV data is given for our programme stars.
In this figure relations for luminosity classes III and V are plotted using
the intrinsic 
and 
colours listed in
Schmidt-Kaler (1982).
Furthermore, the reddening slope of E(U-B)/E(B-V) = 0.72
is indicated. This value corresponds to the normal extinction law,
characterized by 
: 
.
Such an extinction law has also been adopted for the foreground interstellar
medium ((Thé et al. 1990).
By dereddening the observed UBV data of a star along this line, the intersection
with the intrinsic UBV relations will give us the extinction values, E(B-V)
and E(U-B). The intrinsic values are determined by:
and 
.
Comparing those values to the ones listed in Schmidt-Kaler's (1982) Table 12a
we can derive the photometric spectral type of the star. In this method a
separation in luminosity class seems to be significant only for the coolest objects.
It should be noted that in the above procedure we have used 
= 3.1
for all stars, regardless of their true 
value. This is justified,
because the reddening slope of 0.72 of the UBV system is insensitive to
anomalous extinction, expressed by the 
value.
This can be shown from data of Steenman & Thé (1989 and 1991).
A spectral type derived from the TCD method is, therefore, quite reliable.
For each photometric data set a spectral type was derived.
The ones which are consistent were averaged, resulting in the
spectral type given in Table 4.
As can be seen from Fig. 2 (click here) the photometric spectral type can sometimes not be uniquely derived; there are two or three intersections with the ZAMS possible, resulting in several spectral types for a given star. We will remove this ambiguity using available spectra and the spectral energy distribution (SED) fits, which will be described below.
The Walraven data for our programme stars are listed in Table 1.
Using these data we wish to deduce a photometric spectral type for each
programme star, which can not be simply done as for the UBV data. The Walraven
two-colour diagram is based on the intrinsic colours [B-L] versus [B-U].
The observed colours (B-U) and (B-L) can be
transformed into reddening-free ones using the following
relations (Brand & Wouterloot 1988):
The coefficients in these equations are the slopes of the reddening lines in
the corresponding colour-colour diagrams, and depend on the reddening law
(the normal extinction law has been assumed, i.e. 
, Schmidt-Kaler 1982) and the properties of the passbands. Note that
= 3.2 for the Walraven photometric system is equivalent to
= 3.1 for the UBVRI- Johnson photometric system.
The derivation of log 
and log g from the observed reddening-free
colours is done through linear interpolation in the reddening free
diagrams (Fig. 3 (click here)a) of Brand & Wouterloot (1988).
Then using Table 3 of Schmidt-Kaler
(1982), we find for each star its spectral type listed in Table 4.
Note that the Walraven data are not available for all programme stars
(see Table 1). Note also that several faint and late type stars (for which
this method is not appropriate), W266, W349, W402, W494,
W525, W611 and W617, are completely out of the diagram. A few other stars
(W213, W299, W469, W556 and W605) lie too far away from the
intrinsic (
, log g) models. For these stars no reliable
photometric spectral types can be deduced.
For the spectral classification of both the IDS and the CCD spectra of our programme stars we used the spectral catalogue of Jacoby et al. (1984), which contains spectra for classes O-M and luminosity classes V, III and I, and Sect. 4 of Schmidt-Kaler (1982) in which the main spectral descriptions are given for the MK classification.
A complete description of the main criteria that were applied to classify the available spectra of our programme stars is presented by Koulis (1993). The reduced IDS spectra are presented in Appendix A of Koulis (1993). The reduced CCD spectra are given in Fig. 1 (click here). All deduced spectral types are collected in Table 4.