A star is classified as chemically peculiar (CP) if its spectrum shows
abnormal line strengths of one or several elements. The chemically peculiar
stars were divided by Preston (1974) into four subgroups according to their
peculiar spectral features: CP1 (Am stars), CP2 (classical
magnetic Ap stars), CP3 (HgMn stars) and CP4 (He-weak stars). Subsequently,
it was shown that some CP4 stars were related with the CP3 stars,
while others were related with the CP2 stars and new groups were identified:
the He-rich and the 
-Boo stars.
The distribution of degrees of peculiarity does not show any gap between normal and CP2 stars, instead the smooth transition between normal and peculiar stars forces CP-investigators to set a threshold value (e.g. 3 s.d. of normal stars) for all of the related peculiarity parameters in order to distinguish peculiar stars.
The range of temperatures for CP phenomenon is about
(Lanz 1993) where the stellar atmospheres are stable: the
lower limit corresponding to temperatures where convection is not yet relevant
whereas at the upper limit radiatively driven stellar winds start to
cause significant mass loss.
CP2 stars, including both Bp and Ap, differ from other
peculiar stars in that they possess strong global magnetic fields
(Lanz 1993). These stars are main-sequence objects
and, as pointed out by North (1993), they do not
undergo great changes during their
life on the main sequence. Mechanisms involving diffusion
produce anomalies in non-evolved stars yielding a separation of chemical
elements in different layers at the surface of
magnetic stars (Cowley 1993). As a result, CP2 stars exhibit three
main broad band flux depression features centered on
, 
and 
(Kodaira 1969). The depression at ,
with a two component structure, extends over
according to Adelman (1980).
This depression is at its maximum at temperatures of about
(Hauck & North 1982).
The feature at is also characteristic of CP2 stars
but it is not present in all of them. Finally,
a still smaller number of CP2 stars present the flux
depression.
Adelman & Pyper (1993) described a feature at 
probably
due to abnormally strong line absorption in the Balmer continuum.
Although generally yielding much less diversity of astrophysical
information than spectroscopic techniques, photometry has been increasingly
used for the identification of CP stars since one can obtain data on a
much larger number of stars than is possible through spectroscopy.
Moreover, with photometry a much fainter limiting magnitude is
attainable and the photometric indices are free from personal judgment,
unlike the spectroscopic classification.
While broad band photometric systems, like the UBV system, are ineffective in
detecting the features of CP2 stars,
the photometric indices of intermediate and narrow band systems are perceptibly
modified by the peculiarities.
The most effective detection
is obtained with filters specifically designed to measure
the characteristic features in the spectrum, such as the
-system (Maitzen 1976).
The Geneva system, in spite of being designed for more general purposes,
has proved its capability to identify CP2 stars
(North & Cramer 1981).
In the 
system, usually only the m1 index is used to
discriminate CP stars, although several earlier studies
(Cameron 1967;
Hauck 1975) have shown that the peculiar features in the flux distribution
alter all the Strömgren-Crawford indices.
The aim of this paper is to harness this to
optimally separate CP2 stars from normal stars, finding
the best combination of colour indices through the
Multiple Discriminant Analysis.
We consider different temperature regions for which individual definitions
of our peculiarity parameter are provided.
The determination of physical stellar parameters,
such as effective temperature, surface gravity,
absolute magnitude or age, needs a previous criterion of classification of
stars, since the calibrations are different for normal and CP2 stars.
The peculiarity parameter defined in this work provides this
classification and is especially useful for those stars with only
measurements.
Moreover, it provides a list of CP2 candidates among those
stars with already measured indices.
We also compare the colour excesses computed for early CP2 stars from Johnson and Strömgren-Crawford photometries and give a correction to the reddening computed assuming these CP2 stars to be normal.