5 $\mathsf {H\alpha}$ photometry

The H$\alpha$ photometry we made was used to find stars showing H$\alpha$ emission. This was done by means of an $\alpha$ index, defined as a difference between the magnitudes of a star through a narrow and a wide filter. Our filters and, consequently, the $\alpha$ index resembles those used by others, namely Abt & Golson (1966), Tebbe (1969), Feinstein (1974), Dachs & Schmidt-Kaler (1975), Claría & Escosteguy (1981), Strauss & Ducati (1981), and Goderya & Schmidt (1994). We made, however, no attempt to transform our $\alpha$ to that defined by others; it was left in the instrumental system. In Fig. 4 we show the $\alpha$ index for 442 stars with $R_{\rm C} <$ 15.4 mag, the limiting magnitude of our H$\alpha$ photometry. The same stars are shown in Fig. 5, where the $\alpha$ index is plotted as a function of $(R-I)_{\rm C}$.

Out of the 25 known Be stars in the area shown in Fig. 1, 22 fall within the P field, and 14 within the V field. As can be seen in Fig. 4, except for doubtful Be star Sanduleak 11, all these 22 known Be stars indeed show emission, that is, have $\alpha$ placing them above the locus of cluster non-emission stars, shown with a continuous line. The line was obtained from the spectra given by Danks & Dennefeld (1994) which were combined with the transmission curves of the H$\alpha$ filters provided by the manufacturer. Except for known Be stars, some other stars are found above this line as well. Because some foreground late-type stars can also have weak emission, we need to separate them from the Be stars first. For this reason, we made a rough selection of non-members with the use of the cluster CM diagram (see Sect. 7). Stars selected as non-members are plotted with crosses in Figs. 4 and 5. Obviously, with this kind of selection, we cannot indicate all non-members which contaminate the cluster main sequence. Many objects with $R_{\rm C} >$ 14 mag in Fig. 4 situated clearly above the continuous line are certainly such unrecognized non-members.

Using Fig. 4 we can conclude that four stars, namely W61, 128, 151, and 240 (we precede star numbers taken from Wallenquist (1929) with "W''), may be classified as new Be stars. A star was selected as a new Be star if: (i) the difference between the observed $\alpha$ and the value taken for a star of the same magnitude showing no emission (continuous line in Fig. 4) was negative and its absolute value was larger than 4 $\sigma_\alpha$, where  $\sigma_\alpha$ is the rms error of $\alpha$, and (ii) the star was not selected as a non-member.

Figure 4: The $\alpha$ index for the 442 stars brighter than $R_{\rm C}$ = 15.4 mag, plotted as a function of $R_{\rm C}$. The symbols denote: crosses, non-members; open circles, known cluster Be stars; encircled dots, newly discovered Be stars; filled circles, remaining stars. Known Be stars are labeled with their Sanduleak (1979, 1990) numbers. The dashed line corresponds to the expected value of $\alpha$ for zero total equivalent width of the H$\alpha$ line. The continuous line is an average relation for main-sequence non-emission B and early A-type members. At the bottom, the spectral types for the cluster main sequence stars are given.

Figure 5: The $\alpha$ index plotted versus the $(R-I)_{\rm C}$ colour. The symbols are the same as in Fig. 4. For clarity, error bars and labels were omitted.

All four new Be stars are among the faintest Be stars in NGC663, implying that the previous Be star surveys were rather complete for the brightest cluster stars. It can be also seen in Fig. 4 that Be stars with strong and intermediate H$\alpha$ emission occur in the whole range of the B spectral type. Most of them, however, occupy the range between B0 and B3. Maeder et al. (1999) compared the fractions of Be stars in clusters of an age comparable to that of NGC663. In the widest interval considered by these authors, -5 < M_V < -1.4 mag, they found that NGC663 contained 34 $\pm$11% of Be stars. The above-given interval of M_V corresponds to $8.7 < R_{\rm C} < 12.2$ mag. There are 59 stars we observed in this interval of $R_{\rm C}$, including 18 Be stars. This gives the revised fraction of Be stars in NGC663 equal to $31 \pm 8$%, with the error calculated assuming Poisson statistics.

Figure 6: The $I_{\rm C}$ light-curves of 10 Be stars from the V field of NGC663. The ordinate ticks are separated by 0.1 mag. Note that the time axis is not continuous.

Among stars fainter than $R_{\rm C} = 12.2$ mag, we find 7 Be stars, including all four discovered in our search. Comparing this number with the number of all B-type stars later than B3 (101), we obtain the fraction of $7\,\pm\,3$%.

For 13 Be stars, simultaneous (in the sense that a narrow H$\alpha$frame was sandwiched between two wide H$\alpha$ ones) H$\alpha$photometry was made in one epoch only. For the remaining 13 Be stars, such observations were made on two different epochs allowing detection of possible changes in the emission strength. The largest change of $\alpha$ we detected was 0.045 mag for Sanduleak 3 and two epochs separated by about 550 days. The epochs of H$\alpha$ observations and corresponding $\alpha$ values are given in Table 3, available in electronic form only.

The $\alpha$ indices given in Table 1 were calculated using all frames, including some additional wide H$\alpha$ frames made on epochs other than those given in Table 3. For this reason the $\alpha$ indices given in Table 1 (average values) are not exactly the same as those presented in Table 3 (epoch photometry).