4.1. The lightcurves
We have detected a bump in the B and () lightcurves very close to the secondary X-ray maximum (see Fig. 5a; see also Fig. 2), and a brightening in the V band (and possibly also in B and R) close to the tertiary X-ray maximum at 200 days after the peak.
It is noteworthy that these minioutbursts are different from those shown by MM Vel (Bailyn & Orosz 1995) and by V518 Per (Chevalier & Ilovaisky 1995) because the latter objects show purely optical 'reflares' well after the end of the X-ray outburst which have no X-ray counterpart.
From Table 3 we also note that the decay is slower as we move to higher wavelengths: this may be due to the cooling of the X-ray illuminated zones of the binary system (i.e., the outer disk and the inner face of the secondary).
Other longer-term light fluctuations, observed during the decline, might be real and correspond to faint secondary maxima. In particular, a sort of 10-day periodicity appears to be present during the decline in the B lightcurve (Fig. 5a). According to Warner (1995; and references therein), this behaviour is similar to that shown by some classical novae during the transition phase.
4.2. The spectra and the disk stability
During the first five days of the decline the EW of H drops quite steeply by a factor 2.5, then it remains at about the same value throughout the first half of 1991.
The He II line shows a bump which approximately starts at JD 2448340, in coincidence with the secondary maximum of the X-ray (Kitamoto et al. 1992) and the B lightcurves. The increase of luminosity in the high-energy bands (UV and X-rays) should then be responsible for this bump, since the strength of this emission line is actually correlated with the UV continuum level (Garnett et al. 1991). Later on, the He II line becomes weaker, as also shown by Cheng et al. (1992; their Fig. 3c).
The EW of the 4640 blend appears slightly stronger than that of He II before JD 2448340 and it seems to remain at about the same level throughout the brightening of the He II component. It shows no increase when the X-ray and the B lightcurves have a secondary maximum (see Fig. 2 and Fig. 5a,b). This behaviour could be explained by the rather broadened and undefined profile of the 4640 blend and/or by its lower excitation potential with respect to that of He II.
The FWHM (Fig. 5c) of H and of He II tend to increase with time. The FWHM of the 4640 blend starts with larger values, and then it seems to decrease. The widening at JD 2448304 shown in Fig. 5c is rather uncertain due to the difficult determination of the profile. The FWHM's of He II and of the 4640 blend seem to decrease beginning on May 1991. Anyway, the lack of spectroscopic observations after May 19, 1991 does not allow us to fully confirm this trend. This fact would indicate that during the decline the emission region moves inward in the disk, that is, towards larger keplerian velocity radii.
However we note that X-ray novae may present opposite behaviours. Some objects, like V518 Per (Shrader et al. 1994) and GRO J1655-40 (Bianchini et al. 1997), show emission lines with decreasing widths with the time; some other ones, like GU Mus (this work), V404 Cyg (Gotthelf et al. 1992) and V616 Mon (Whelan et al. 1977) show emission lines which become larger and larger during the decline.
To investigate this point we plot in Fig. 6 the ratio FWHM /FWHM of the H line, at about 150-200 days after the X-ray peak and at maximum, versus the X-ray luminosity at maximum of the 5 SXTs with available data (in Table 5), i.e. GU Mus (this work), V404 Cyg (Gotthelf et al. 1992), V616 Mon (Whelan et al. 1977), GRO J1655-40 (Bianchini et al. 1997) and V518 Per (Shrader et al. 1994). The errors on Log(L ) are mainly due to the uncertainty on the distance of the objects (see Table 5), while those affecting FWHM ratios are originated by the signal-to-noise ratio of the spectra. The long-dashed line represents the weighted least-squares fit of the data points:
The correlation coefficient of the data is 0.96, which indicates (admittedly on the basis of a scanty statistic), that these quantities seem to be correlated.
Table 5. Optical decline rates, distances, time from the X-ray peak at which the FWHM of H was measured and corresponding magnitude drops ( m) for the 5 SXTs quoted in Sect. 4.2
The regression line of Fig. 6 suggests a peak luminosity erg s-1 for the case FWHM /FWHM =1. This value of the luminosity approximately represents the Eddington luminosity of SXTs. We note that the behaviour of the sub-Eddington objects would typically represent the one it is expected from the outburst caused by an enhancement of mass transfer from the secondary. Actually, former disk instability event has only triggered the mass transfer paroxism. In fact, during the early outburst the disk becomes smaller because of the very large flow of low angular momentum material from the X-ray heated secondary; so we may observe larger emission lines. During the decline the disk relaxes and the emission region will expand again towards lower keplerian velocities as it was suggested for GRO J1655-40 by Bianchini et al. (1997).
For super-Eddington objects, like GU Mus, accretion from the inner regions of the disk should be inhibited and the disk itself might be partially disrupted. More likely, we might argue that at super-Eddington accretion luminosities the disk turns its structure into a 'geometrically thick' one (see e.g. Frank et al. 1992), in which only the outer cooler regions can be seen at almost all inclinations. In this case, at light maximum the line emitting region is placed at quite large radii, thus producing narrower emission lines than during the decline, when a 'standard' disk is formed again.
© European Southern Observatory (ESO) 1998
Online publication: December 8, 1997