6 Cross-identifications

The detected variable stars are uniformly distributed along the CCD chips, and no obvious bias (due e.g. to internal extinction) are suspected. In order to appreciate the sensitivity of our analysis, we cross-identified this catalogue with previous works. Among the 631 variable stars presented here, 72 have already been studied elsewhere. We considered the main catalogues which covered this area, namely the catalogues of LT$\&$LPV by Hughes (1989), eclipsing binaries and cepheids by EROS (Grison et al. 1994 and Beaulieu et al. 1996), radio sources (Marx et al. 1999), X-ray sources (Haberl & Pietsch 1999), as well as the General Catalogue of Variable Stars (Artiukhina et al. 1995) and the NED data-base. We define a genuine cross-identification any object whose offsets with respect to the variable objects detected here are smaller than $\Delta \alpha_{\rm max}=5\hbox{$^{\prime\prime}$ }$ and $\Delta \delta_{\rm max}=5\hbox{$^{\prime\prime}$ }$.

Figure 5: Histogram of the periods. We present here the periods measured elsewhere which correspond to 57 of the 631 variable stars detected here

Among the 58 LT$\&$LPV found by Hughes (1989) in this field, 41 have been detected here, 12 are missed due to CCD defects (one of those is rejected with the visual inspection described in Sect. 3.3). This gives the order of magnitude of the completeness of our sample over a 120-day window: we typically miss up to 20$\%$ of genuine variable stars due to CCD defects. More interestingly, the last 5 missing LT$\&$LPV escape our selection for no obvious reason. One of them (SHV0527122-695006) has been subsequently observed in the near-infrared (Hughes $\&$ Wood 1990): it did not follow the (K, $\log P$) relation, and was interpreted as a supergiant or a foreground Mira. According to the DAOPHOT magnitudes measured at epoch JD = 2448678.3 $B_{\rm EROS}=15.7$ and $R_{\rm EROS}=13.7$, it is consistent with a supergiant. The 4 remaining, namely SHV0516251-693241, SHV0519415-693441, SHV0522220-694441, SHV0522251-692902 - not observed subsequently by Hughes $\&$ Wood (1990) - were reported by Hughes (1989) with a low amplitude ( $\Delta I \simeq 0.5$ mag) and 3 of them were only marginally periodic. Moreover, our non-detection can be explained for some of them by changes in mass-loss rates as suggested by Whitelock (1997). The extension of this work with a longer baseline together with DENIS photometry is expected to provide further arguments about this kind of behaviour (in preparation).

The short periodic variable stars from the EROS catalogue (Grison et al. 1994; Beaulieu et al. 1996) are missed here for most of them: short timescale variations are broken by the averaging procedure as explained in Sect. 2. Only 22 out of the 181 variable stars previously detected by EROS are present in this catalogue. In addition, 5 of the 7 pre-main-sequence candidates published by Beaulieu et al. (1996) have been detected. None of the radio-sources, detected by Marx et al. (1997), are lying in the studied field. 2 X-ray sources detected with ROSAT (Haberl & Pietsch 1999) are present in the field, but none of our variable stars lies in the 90$\%$ CL error box of those sources. 29 extragalactic sources (but not necessarily variable) from NED are present in the field, none is among the variable stars detected here. Last, whereas 129 sources from the GCVS are lying in the field, 52 are selected here. Most of them were also in the catalogues mentioned above. According to the GCVS classification, among these 52 variables, 17 are Miras, 26 semi-regulars, 2 irregulars, and 7 cepheids.

A total of 72 sources out of 631 were previously reported, 57 of which have a published period. Those are presented in Fig. 5: they are clearly of the order of 100 - 300 days. A visual inspection confirms that besides a few short periods, these timescales are quite representative of the bulk of a distribution dominated by LT$\&$LPV.

Figure 6: Examples of red variables. The light curves in red and blue together with the corresponding finding chart in red are displayed for 4 red variable stars from the catalogue. The $\phi _{\rm c}/\phi _0$ ratio is provided for each light curve, as well as the magnitude at JD = 2448678.3. Note the fact that the large dispersion with respect to the errors observed for some blue light curves is due to blending and those light curves have a low $\phi _{\rm c}/\phi _0$ ratio

Figure 7: Examples of red variables. Same as Fig. 6