3 Geneva photometry

The Henize sample of S stars has been monitored in the Geneva photometric system on the Swiss telescope at La Silla (Chile). Detailed information on this photometric system and on the data reduction can be found in Golay (1980), Rufener (1988) and Rufener & Nicolet (1988). Among the 205 S stars of the Henize sample, 179 could be reached with the 70 cm Swiss telescope, with an average of 4 good-quality photometric measurements (in all filters) per star.

3.1 Rejected measurements

Some photometric measurements have been discarded. They concern mostly misclassified stars, or stars with abnormal colours:

• Misclassified stars: Hen 22 and 154 are two stars misclassified as S, as revealed by low-resolution spectra (Van Eck & Jorissen 1999, hereafter Paper I; see also Paper III). Hen 22 is a mid-K giant ($\sim$K3-5) and Hen 154 a late-G giant ($\sim$G8);

• Single measurement revealing abnormally blue colours: Only a single very blue measurement (U-B=2.26; B-V=0.48) is available for Hen 82; a nearby field main-sequence star could well have been measured instead of the faint and variable S star Hen 82. In fact, the S star Hen 82 is most probably a Mira (strong H$_\alpha$ and H$_\beta$ emission are visible on its low-resolution spectrum; spectral type S5,8 according to Henize (as listed in Stephenson 1984). Although Mira stars are known to sometimes exhibit such blue colours, Hen 82 has been removed from the photometric data set, given the lack of additional photometry confirming the blue colours of this star. The same holds true for Hen 181 (=V407 Sco; U-B=2.17; B-V=-0.53);
• Close visual binaries: Several stars (Hen 47, 94, 105, 155) were found to be close visual binaries, and their photometric data have therefore not been considered for the cluster analysis presented in Paper III.

  In all these cases, two spectra separated by less than 10'' (angular separation on the sky projected on the direction perpendicular to the slit, corresponding to less than 15 pixels on the CCD frame) have been recorded by the Boller & Chivens spectrograph (Sect. 7). Rough estimates of the companion spectral type could be obtained in all cases, but Hen 94 which is too faint: Hen 47 (>5''): S + K4V; Hen 105 (>4''): S + late BV or early AV; Hen 155 (>14''): S + G5 IV-III. The numbers in parentheses provide a lower limit on the angular separation.

  Hen 47 is the only one among those cases where the physical association between the two stars is clearly unlikely. The difference in absolute visual magnitudes between the S star and a K4V star is at least 7 magnitudes, but the observed difference in apparent visual magnitudes is less extreme, since the available photometry clearly reveals composite colours.

  All these pairs lie within $2^\circ$ of the galactic plane (except Hen 155 with $b = 20^\circ$) where crowding may be severe. If some of these pairs would nevertheless turn out to be physical, the less evolved main sequence companion would allow to set a lower limit on the mass of the S star, namely >3  $M\sb{\odot}$ for Hen 105. These masses are compatible with S stars being intermediate-mass stars;

• Magnitudes fainter than 16.5: All magnitudes fainter than 16.5 were rejected, as well as their associated standard deviations, because at such faint magnitudes the flux contribution from the sky is dominating. For our S stars, magnitudes fainter than 16.5 only occur in the U filter, except for Hen 122 (B = 17.82), which is flagged in Table 4.

  The stars having U>16.5 fall in two categories: (i) either the star has some of its U measurements fainter than U=16.5; in that case only the measurements brighter than U=16.5 were considered; (ii) the star has all its U measurements fainter than U=16.5; no U standard deviation is computed in that case. The column labelled n_U in Table 4 indicates the number of measurements fainter than U=16.5.

3.2 Average colours and standard deviation

The average magnitudes were computed from the weighted fluxes according to Eq. (2) of Rufener (1988), where the weights refer to the quality of the measurements.

The reduced standard deviation $\sigma_{\rm r}$ is computed by quadratically subtracting the instrumental error $\sigma_0$ from the standard deviation $\sigma$. The instrumental error is computed by interpolating Fig. 2 of Barblan et al. (1998), giving the mean precision as a function of the magnitude for observations performed in the Geneva system. If $\sigma\le\sigma_0$, then $\sigma_{\rm r}=0$. The reduced standard deviation has not been computed for magnitudes fainter than 16.5, because of the large uncertainties affecting the mean precision $\sigma_0$ at such faint magnitudes.

   
3.3 Dereddening

Each star has been dereddened according to the following procedure:

(i) If $10^\circ\le\vert b\vert<45^\circ$, the colour excess E_B-Vis taken from Burstein & Heiles (1982) and is multiplied by the factor $[1- \exp(-10~r~\sin\vert b\vert)]$, where r is the distance in kpc and b is the galactic latitude (Feast et al. 1990); 72 stars are concerned. The visual extinction is then computed with $A_V = R \times E_{B-V}$, where R=3.1.

(ii) If $\vert b\vert<7.6^\circ$, A_V(r) is taken from Neckel & Klare (1980); 70 stars are concerned.

(iii) If $\vert b\vert\ge 45^\circ$, or if the star is not on the Milky Way fields defined by Neckel & Klare (1980), or if it falls outside their A_V(r) diagram (i.e., if the star is located too far away), then the visual extinction is taken from Arenou et al. (1992); 37 stars are concerned.

The dereddening procedure then requires to assign absolute visual magnitudes to S stars. From HIPPARCOS parallaxes, intrinsic S stars are known to be brighter than extrinsic S stars (Van Eck et al. 1998), but several factors (intrinsic variability, as well as the Lutz-Kelker bias) prevent from giving accurate average luminosities for both classes of S stars. The only previous large-scale estimate of absolute visual magnitudes of S stars is by Yorka & Wing (1979), who derived that the average M_V at maximum light is of the order -1.5 to -2.0 for Mira S stars (i.e., intrinsic S stars) and -1 for non-Mira S stars (presumably mostly extrinsic S stars).

In principle intrinsic S stars and extrinsic S stars should thus be assigned different absolute magnitudes (say M_V=-1 for extrinsic S stars and M_V=-2 for intrinsic S stars), depending on whether they have technetium or not. This approach has not been retained here since (i) the only unambiguous way to distinguish extrinsic from intrinsic S stars is technetium detection, and this information is available only for 70 S stars (out of 205); the other parameters capable of segregating the two kinds of S stars have only a statistical efficiency (see the discussion about Sb in Paper III), and (ii) this would introduce an a priori distinction between extrinsic and intrinsic S stars, which would then be difficult to disentangle from possible genuine photometric differences derived subsequently.

While the intrinsic S stars are brighter than the extrinsic ones, they are also much redder, hence their V magnitude is dimmed. Therefore the single plausible value of M_V=-1 has been assigned, for dereddening purposes only, to both extrinsic and intrinsic S stars.

The dereddening process is iterated until convergence of the apparent V magnitude is achieved (to a level of 10^-4 mag). The dereddened U and B magnitudes are then computed, using $E_{U-B} = 0.652 \times E_{B-V}$, a relation derived by Cramer (1994) for B stars in the Geneva photometric system (no such relation is available for late-type stars).