Online material

Appendix A: Obtaining the O–C diagram

	Fig. A.1 Amplitude periodogram of the light curve shown with all the eclipses removed and interpolated. The vertical dashed lines indicate the period of the close binary 1 and its harmonics. The red ticks indicate the position of the frequencies removed for further analysis.
Open with DEXTER

Fig. A.2 Averaged curves of the primary and secondary eclipses of close binary 1 and best-fit trapezoid model to each of them. The points around 0.968 are the residuals, shifted vertically for visualization. There are significant differences from the simple trapezoid model, but as these are approximately symmetrical from the centre of the eclipses, they are not expect to introduce systematic errors in the determination of the time of eclipse centres.

Open with DEXTER

To prepare the light curve for the analysis, first we cleaned the most significant frequencies in the amplitude spectrum of the off-eclipse sections, using period04 (Lenz & Breger 2005). Twelve sinusoidal components were used, until amplitudes of the peaks of about three times the dispersion of the residuals

were reached. Figure A.1 shows the amplitude spectrum and the frequencies that were removed. As in the case of the autocorrelation, the spectrum shows frequencies that are close to, but not exactly, that of binary 1; the spectrum also shows peaks at the orbital frequency of the binary 1 and its harmonics. These might be caused by a combination of reflected light, ellipsoidal variability, and doppler beaming whose detailed analysis is beyond the scope of this paper.

We used a parabolic fit to the baseline to remove the modulation between the eclipses, and a Levenberg-Marquardt fit (Press et al. 1992) to a trapezoidal function to estimate the parameters of each eclipse (centre, depth, duration, and ingress time). Of the four fit parameters for each eclipse, only the central time of the eclipse shows a significant variability, while the other three remain constant. The observed variability has the same period as the wide component A. On a final iteration, we corrected for the measured O–C displacement at each time stamp before computing the average curves for both the primary and secondary eclipse of the inner binary (templates), which are plotted in Fig. A.2. For each observed eclipse, the corresponding template was shifted in time on a regular grid and the χ² between the model and the data was calculated and used to calculate the time of eclipse centre and its uncertainty.

Appendix B: Additional tables

© ESO, 2015