3.1. Total flux monitoring
YZ CMi was found at a surprisingly high, slowly decreasing flux density (average 2.9 mJy) during the whole observation (see Fig. 1, left plot). As previously noted (see Sect. 1), the emission of YZ CMi is often considerably polarized. Throughout this observation the polarization was predominantly left circular (about 60% circular polarisation during the quiescent emission, and 90% during the flares). Two strong flares appear in the lightcurve obtained with the VLA. Their flux values reach up to 6.0 mJy. The detection of YZ CMi with both instruments was clear, as the rms noise in the VLA image was 0.063 mJy/beam, and in our high sensitivity VLBI maps made using only VLBA-VLA baselines the noise was 0.13 mJy/beam. Given these map noise values the peak intensity of the star was 47 in the VLA image, and 16 in the VLBA image.
The mean flux density value for AD Leo was at a more typical level of 0.7 mJy (see Fig. 1, right plot). Similar fluxes at 18 cm have previously been reported (e.g. Jackson et al. 1989; Benz et al.1995). A weak flare appears in the VLA lightcurve (2.1 mJy). The star was detected by both instruments despite its low flux. The rms noise for the VLA map was 0.035 mJy/beam, while for the map made with only the VLA-VLBA baselines it was 0.18 mJy/beam. Peak map values were 15.6 for the VLA and 5 for the VLBA maps respectively. Tracing the lightcurve of AD Leo from the interferometric VLA data was a delicate task because of two strong sources in the field of view.
3.2. Size and shape of YZ CMi
The phase referenced VLBA map of YZ CMi is shown in Fig. 2. Evidence for spatial resolution was found in this map. Further evidence for extended emission can be most clearly seen by examing fringe amplitude versus baseline length obtained using the VLA-VLBA baselines (see Fig. 3, left plot). We obtained this plot by first coherently averaging the data on each baseline-scan (75 minutes) in
order to increase the signal to noise ratio significantly above unity. We then binned the amplitude over baseline length finding the mean amplitude in each bin by incoherent averaging. The error bars on this average were determined from the internal scatter of the data.
The fall off noticeable in Fig. 3 left plot clearly indicates a resolved source. What is more, the phase values on VLBI baselines to the phased VLA over the whole observation show no significant variation from zero. There is therefore no evidence for anything other than a single centro-symmetric component. We searched also for other evidence for non-symmetrical structure looking at closure phases, but the SNR of these were too low.
Given the close to quadratic fall off of the fringe amplitude with u,v -distance shown in Fig. 3 left plot it is impossible to distinguish between gaussian, sphere, disk or ring like models (Pearson 1995). We therefore fitted one-component gaussian models to the YZ CMi data. The numerical values for the fits are summarized in Table 1. The dimensions of sphere, disk and ring which would show similar fits are expected to be respectively 1.8, 1.6 and 1.1 times the gaussian FWHP values (Pearson 1995). Two ways of fitting were followed: fitting in AIPS using the task UVFIT and our own model fitting to the data. The first fitted an elliptical gaussian and obtained for the whole data set a major axis of FWHP of 1.4 0.3 mas and a minor axis of 0.5 0.25 mas (1 errors). There is therefore no strong evidence for ellipticity, and our subsequent fitting of the data outside of AIPS assumed only a circular gaussian. With such a model it was possible to fit most of the data within 1 . The best fitting FWHP size of the corona was found to be 0.98 0.2 mas, which corresponds to 1.7 0.3 stellar diameters.
Since the second VLBI scan corresponds precisely to one of the two strong flares (see Fig. 1, left plot), it was interesting to study it more closely. Fig. 3 (right plot) shows the amplitude versus u,v -distance. Of this scan we selected the subscan on the target with the highest flux density value and coherently averaged the data over it (3 minutes), obtaining one point per VLA-VLBA baseline. The solid line corresponds to the same gaussian model fitted to the whole data set (see Fig. 3, left plot) except for the total flux density which was increased to 5.1 mJy. We find that this scaled model fits the data within the errors and therefore there is no evidence for a change in source size during the flare. Independent gaussian fits to the data are also consistent with this conclusion (see Table 1).
We should add that the contribution of the proper motion and of the changing parallax of the star during the ten hours of observation is -0.32 mas and -0.16 mas in and , respectively. These values are small enough that they do not give a significant contribution on the spatial extent.
3.3. AD Leo
An image of AD Leo is shown in Fig. 2. It appears slightly elongated. This image is probably affected by the star's high proper motion (Table 1), since the extension is exactly along the expected direction: the star appears blurred because of its motion during the synthesis observation.
As shown in the flux curve of AD Leo (Fig. 1, right plot), the total flux varied by a factor of 3 during the observations. Adding this fact to the high proper motion of the star, it is not surprising that attempts to fit a single gaussian to the -data did not give consistent results. The weakness and variability of the star make the estimation of errors on the size difficult. However, from the image (see Fig. 2), we can note that the apparent FWHP perpendicular to the motion corresponds to the beam FWHP in this direction. The intrinsic FWHP size of the emitting region is therefore likely to be less than half the beam FWHP or about 1 mas, which equals the estimated optical diameter of the star (see Table 1). This might indicate a very compact corona or an emitting spot on the surface of the star. A very conservative upper limit on the size of the corona in the former case would be to assume the emitting region to be an optically thin sphere instead of a gaussian (Pearson 1995) in which case the diameter is less than 1.8 times the photosphere diameter, and it therefore has an extent above the photosphere of less than 0.8 .
The relatively large signal-to-noise ratio for YZ CMi and the phase referencing to a calibrator with good (0.5 mas) positional accuracy in the radio frame (Johnston et al.1995) allowed us to determine a precise position for this star. This position was compared with the position given by the Hipparcos catalogue (ESA 1997). Correcting for proper motion and parallax, we found a discrepancy of 20.9 mas in and 30.4 mas in , thus a total deviation of 36.9 mas. The proper motion of the star is given in the Hipparcos catalogue as -344.9 2.6 mas/yr in , and -450.8 1.75 mas/yr in . Considering the time interval between the two measurements of 6 years, the difference is 2 and thus within the accuracy of the proper motion error bars. Combining the VLBA and Hipparcos positions (courtesy of F. Arenou), an improved proper motion of -348.6 0.6 mas/yr in , and -446.6 0.3 mas/yr in can be derived.
The position of AD Leo obtained with the VLBA was compared with those available in the Gliese and Tycho catalogues. The latter showed a deviation with the VLBA position of 176.3 mas and 100.0 mas in and , respectively. They are within one standard deviation of the Tycho catalogue accuracy.
© European Southern Observatory (ESO) 2000
Online publication: December 17, 1999