2 Observations and data processing

We observed with the VLA in its most extended A configuration at the 6 cm wavelength, equivalent to a frequency of $\nu=4.86$ GHz. Three ToO VLA runs of 5 h each were scheduled at two-week intervals starting on late October 2000. The data were reduced using the AIPS package of NRAO. The source 3C286 was observed as amplitude calibrator with the range of baselines restricted to 150-300 k$\lambda$ and its nominal flux density reduced by 2%, as recommended in the AIPS CookBook. The phase calibrator used was 2007+404 at a distance of 4.7 degrees. The log of observations and the bootstrapped flux densities of 2007+404 are listed in Table 1.

The observing epochs are indicated in Fig. 1 as vertical arrows overlapped with the 8.3 GHz monitoring provided by the Green Bank Interferometer (GBI). The series of outbursts that triggered the activation of our ToO proposal are clearly visible to the left of the arrows. Consultation of the 15 GHz radio monitoring provided by the Ryle Telescope, kindly posted on the web by G. G. Pooley, indicates that Cygnus X-3 was at a quiescent level of radio emission throughout all the epochs of VLA observation. The vertical dashed line labelled as ejection corresponds to 15 September 2000 (JD 2451802.5), the approximate date when the radio outburst activity started. A very strong outburst event in April 2000 (not shown in Fig. 1) did not trigger our observations because of the inappropriate VLA configuration at that time.

  

Table 1: Log of VLA observations at 6 cm.

Figure 1: Radio light curve of Cygnus X-3 as observed at 8.3 GHz by the GBI. A series of strong flaring events are clearly seen reaching up to $\sim$7 Jy at the GBI wavelength. The vertical arrows indicate the three epochs when VLA observations were carried out. The vertical line labelled as ejection corresponds to the starting epoch of the flaring activity that triggered our ToO proposal.

Figure 2: Radio light curves of Cygnus X-3 at the 6 cm wavelength during the three epochs of VLA observations. Each dot is the average of 2 min of visibility data using the AIPS task DFTPL. The source was always clearly variable on time scales of hours. The horizontal arrows and vertical lines represent the data block (or blocks) selected to compute the VLA maps in Fig. 6.

The observation of microquasars with radio interferometers is very often made difficult by the brightness and/or structure changes occurring on time scales of a few hours, i.e., comparable with the typical duration of the observation. The assumption of a constant source, both in brightness and structure, in the commonly used CLEAN deconvolution algorithm are then clearly violated. At the VLA angular resolution, only the fast radio variability of the Cygnus X-3 core may pose problems for standard data processing within AIPS. Such variability is shown in Fig. 2 and it was especially severe during the first run of our observation. In contrast, the proper motion of the ejecta is not expected to introduce any noticeable changes in the source structure throughout the run duration.

The variability problem in Cygnus X-3 has been addressed by Martí et al. (2000) by subtracting a variable point source with the estimated light curve of the core. When applying this method to their 1997 data, it was possible to clearly see the faint jet-like arcsecond features that otherwise would be masked by the variability artifacts. As it will be seen later, the Cygnus X-3 ejecta in late 2000 were not well detached from the core when we observed with the VLA. Therefore, the Martí et al. (2000) procedure is not straightforward to apply due the difficulties in estimating a good light curve for the core. Fortunately, Cygnus X-3 behaved in relatively stable way during a significant fraction ($\sim$2 h) of each 2000 observing run. By splitting and using only the visibility data blocks during these intervals, it is possible to compute VLA maps that are practically free of variability problems. The selected intervals are indicated in Fig. 2 by vertical lines. The amplitude of variation during these intervals was always less than 5%. The visibilities were self-calibrated in phase, using a point source model, prior to performing the splitting.