3. Continuum light curves
Altogether, 19 sets of spectra could be secured, of which 14 were obtained in the course of the monitoring. Most spectra are of very high quality, some with a continuum S/N ratio exceeding 100 in component A. Examples of the spectra are presented in Fig. 1; the range in S/N ratio is bracketed by these example data, with most spectra looking rather similar to the February 1995 ones. The high quality of the data enabled us to monitor the monochromatic continuum fluxes of both components, which has the advantage over the conventional broad-band magnitudes that it is a brightness measure uncontaminated by emission lines and their possibly different variability patterns. However, absolute photometric calibration from standard stars was usually not possible, and we had to design a method to compare spectra taken at different epochs. In the following we motivate and outline this procedure.
After placing the spectra on a relative flux scale, we measured the fluxes and equivalent widths of all major emission lines in both components (Ly, Si IV 1400, C IV 1549, and C III] 1909). Local continua were estimated by fitting straight lines to predefined wavebands known to be largely devoid of emission and absorption lines (cf. examples in Fig. 1). Although a direct comparison of line strengths between different epochs is not possible, there is one strong piece of evidence that the lines have remained essentially constant over the time span observed: The flux ratio between the same lines in components A and B, independent of the absolute scale, has stayed at a consistently constant value of for all lines. This implies either a time delay of much less than a month (the separation between data points during the periods of quasi-continuous monitoring), which is highly improbable, or simply constancy of the line fluxes as such. Adopting the latter hypothesis, we were then able to recalibrate the spectra by scaling them to equal emission line fluxes. As reference we used C IV, a prominent line that is surrounded by clearly identifiable continuum windows visible also at low spectral resolution.
For each pair of spectra we computed a scale factor so that the C IV flux of component A assumed an arbitrary but constant value, and applied this factor to both spectra. The average continuum values in the interval , thus at Å, were then determined for A and B. This estimate of the QSOs brightness is independent of external flux standards and of photometric conditions, and we were thus able to incorporate also narrow slit observations by the same method. Note that the relative flux calibration based on standard stars was not even strictly needed, although it helped in removing the instrumentally caused curvature from the spectra. The resulting light curves are depicted in Fig. 2. The error bars in this plot contain both the continuum uncertainties due to photon shot noise and the error of the line flux rescaling factor. The zeropoint for both components is arbitrarily set at the continuum magnitude of A in May 1993, which constitutes the brightest point.
Although the light curve is certainly not well-sampled, some features are very cleary discernible. The strong decline in component A between 1993 and 1995, spanning almost a magnitude, is well mirrored in B, except for the inflection between Nov 1994 and Feb 1995 which occurs only in B. This feature alone is already very suggestive that B leads the variability, as one expects from the observed lens configuration (cf. R98). From early 1996 on, the sampling improved due to the beginning of regular monitoring, and it became apparent that the object shows also significant variability on relatively short time scales.
© European Southern Observatory (ESO) 1998
Online publication: October 22, 1998