4 Non-eclipse variability

The brightness level of the TY$\,$CrA light curve changed significantly during the 5-year period on several time scales. Figure 1 shows that the system became $\sim$$0\hbox{$.\!\!^{\rm m}$}07$ brighter from 1989 to 1992/1993. The 1994 observations (Fig. 2) show that this tendency subsequently reversed, and the system became again fainter than in 1993 by $\sim 0\hbox{$.\!\!^{\rm m}$}03$ in all four colours (see also Fig. 3 in Paper III).

  

Figure 4: Nightly mean O-C values from the theoretical y light curve (Paper III) plotted with all observing periods referred to the first night so as to have a common origin. The night-to-night variations are clearly seen

Non-eclipse variations are also present on much shorter time scales. This is evident in our extensive 1994 data set obtained in an interval covering 42 days ($\sim$14 orbits); the phased light curve is shown in Fig. 2. Variability is evident at numerous phases where the range of photometry can be as large as $0\hbox{$.\!\!^{\rm m}$}05$, far greater than the observational errors. Variability is also evident in the brightnesses of both the shoulders and minima of the primary eclipse. However, we have data from many nights when both the minimum and one of the shoulders of the primary eclipse are observed; in all cases the eclipse depth remains unchanged.

Another perspective on this non-eclipse photometric variations is given in Fig. 4, where we show nightly mean O-C values from the theoretical y light curve computed in Paper III. For the 1994 data there is an evident decrease in the mean light level by $0\hbox{$.\!\!^{\rm m}$}04$ over the course of 30-35 days, followed by a slight increase over the next week. Evidently non-eclipse variations occurs over the course of a few orbits. Occasional significant variations are also seen on time scales of a day, and thus shorter than the orbital time scale. It is clear from Fig. 4 that, in our measurements covering from 1989 to 1994, TY$\,$CrA changed its brightness by $0\hbox{$.\!\!^{\rm m}$}10$ and reached its brightest levels in 1993.

As is evident in the colour-index curves (Figs. 3 and 4 in Paper III), these non-eclipse variations are grey to first order. However, more detailed examination reveals a reddening associated with the brightness variations. Extending the calculation of the mean O-C to the other (bvu) colours, we show in Fig. 5 the mean O-C in (b-y) against the mean O-C in y. There is a clear trend showing that the fainter the star, the redder its colour.

Given this reddening and the lack of phasing with the orbit, it is tempting to associate these light variations with variable dust obscuration, as also has been demonstrated for the PMS binary AK Sco (Andersen et al. 1989), which can show variations of more than 1 mag in the time scale of one day (later unpublished observations). It seems unlikely that long-term variations as large as $0\hbox{$.\!\!^{\rm m}$}05$ are intrinsic to the stars in TY CrA. The secondary and tertiary stars individually contribute less than 3% to the total light, thus making spots on these stars an unlikely origin. Flares are common during the PMS phase of evolution, but they can hardly be the reason for slow variations on time scales of tens of days and longer. An intrinsic variation of the primary star also seems unlikely, considering that the depth of the primary eclipse remains unchanged in the course of the light variations. The same argument applies to varying contributions from the reflection nebula.

  

Figure 5: Colour index curve $\Delta(b-y)$ versus $\Delta y$ for the amount of extinction derived from nightly O-C means using the adopted solution of Paper III. Different observing runs are shown with different symbols. The stronger the effect the redder its influence on the light curves. Mean total-to-selective extinction laws (Cardelli et al. 1989) corresponding to R=3.1 (continuous line) and to R=6.5 (dashed line) are also shown

Barring these explanations, we suggest that at least the longer term variations are due to variable dust obscuration.

Figure 5 shows curves orresponding to mean total-to-selective extinctions of R=3.1 and R=6.5, from Cardelli et al. (1989); R=6.5 is the value found by Cardelli & Wallerstein (1989) for TY$\,$CrA. Our data are also consistent with such a large value of R. Our observations do not allow us to determine the location of the intervening dust. The spectral energy distribution (e.g. Bibo et al. 1992, Paper I) shows a large flux excess longward of $\rm 10\;\mu m$, indicating the presence of circumbinary dust. The increasing flux with increasing wavelength suggests a circumbinary shell, plausibly the source of variable extinction. The presence of a circumbinary disk with a central hole cleared of dust is also possible and may contribute to the variability. For an estimated tertiary orbital radius of order 0.5$\,$AU, disks might form around both the eclipsing pair and the tertiary star as well as surrounding the entire system, and the tertiary star may be able to influence the dynamics of the dust even if it were not to have a disk of its own. A concerted observational effort will be needed to clarify the dust distribution in the TY CrA system in more detail.

However, not all of the photometric variability is straightforwardly explained by circumbinary obscuration. For example, Paper III notes that during 1992 the system became $0\hbox{$.\!\!^{\rm m}$}015$ fainter at phase $0\hbox{$.\!\!^{\scriptscriptstyle\rm p}$}7$ in three sequential orbit cycles, strongly suggesting that the variation was linked to orbital phase, although evidently not to the eclipse phenomena. Undoubtedly, the TY CrA system harbours more secrets than we have yet been able to unravel.