3. Results and discussion
3.1. Integrated spectra
In Fig. 1 we present the Z CMa integrated spectra, i.e., the projection of the data cube on the wavelength axis. The [OI]6300 Å line is visible, with its high velocity wing contaminated by telluric absorption. However, the telluric [OI] emission contribution to the line is negligible. Several FeII lines are also present (Hessmann et al., 1991).
3.2. Cube Gauss fitting and component spectrum recovery
We fitted each monochromatic image with a 2D circular gaussian. This allowed us to estimate a 0.63" FWHM spatial resolution by assuming that these images were unresolved. The precision in the gaussian centroid position was 0.002".
The evolution of the gaussian centroid with wavelength when compared to the spectra revealed that the continuum shifts in the direction of the atmospheric differential refraction - ADR (eg. Filippenko, 1982). The emission lines however have a different behavior: they shift proportionally to their intensity in the direction of the primary. We thus fitted the continuum centroids versus wavelength with a straight line and applied the correction to the cube thus removing the linear ADR drift (the paragalactic angle varied only during the exposure and could be negleted). In Fig. 1 we plot the ADR corrected gaussian centroid position versus wavelength in the binary system frame. All the emission lines except [OI] are associated with shifts in the primary direction without any counterpart in the perpendicular direction. This is due to a lever effect from the primary emission in the fit. The primary is responsible for the emission line component in the integrated spectra and the secondary dominates in the continuum. This is in agreement with the speckle observations, the spectropolarimetric data interpretation and recent observations by Bailey (1998), who measured centroid shifts, from long slit spectroscopy, of Z CMa.
Appendix A demonstrates that the fitted gaussian centroids measure the system barycenter. Hence, our centroid and system spectrum measurements can be combined with the information derived from the speckle data to recover the spectra of each component. Our raw centroid measurements are by definition:
where and are the monochromatic fluxes of the primary and secondary, the system flux, and the primary and secondary positions and the system separation. In the continuum and for our small wavelength range (300 Å) the previous expression does not depend on :
where is the flux ratio of the components in the continuum. The previous expressions can be combined with the information derived from the speckle data (r and ) to recover the spectrum of each component:
where , is the centroid shift, in the system axis, relative to the continuum centroid position (which is computed by taking the median of all points with a shift smaller than the average). If the flux ratio at a given wavelength equals the continuum ratio r we recover . However, the trick of the method is that, because each object has a diferent spectra, locally we have (due to the emission/absorption lines). As a consequence, changes with wavelength as observed in (Fig. 1). Our measurements yield C and (see Fig. 1), using the speckle data in the same wavelength we get and (Thiébaut, 1994, Barth et al., 1994, Thiébaut et al., 1995).
In Fig. 2 we plot the recovered spectrum of both components. The typical (median) error for the primary reconstructed spectrum is 18% (decreasing to 12% in the emission lines). The variance is dominated (65%) by the magnitude ratio error which we took as 0.2 mag. The secondary spectrum reconstruction error is 3%.
The primary is the emission line object. The spectrum of the secondary presents several absorption features that were visible in the low state system spectrum. The broad absorption complex seen at Å was present in the low state spectrum - Fig. 2 of Hartmann et al., 1989 and no sign of it is present in the system integrated spectrum presented in our Fig. 1. The absorption features seen at Å and at Å and already hinted in the integrated spectrum were also present in Welty et al., 1992 low state spectrum (their Fig. 3).
3.3. The [OI]6300 jet
The [OI] emission in the reconstructed spectra shows a more complex behavior. In Fig. 1, the centroid shifts at the [OI] position move away from the secondary towards the primary. This shows that the primary drives the jet. Furthermore there is also a significant shift in the direction perpendicular to the system axis. This shift is in agreement with Bailey (1998), who found that in his long slit centroid measurements the [NII] and [SII] emission lines profiles are dominated by the jet. These [OI] shifts in the perpendicular direction imply that the reconstructed spectra in the [OI] region are not correct because we assume that all the emission is concentrated in the binary. Furthermore they hint a more complex structure for the [OI] emission region.
We integrated spectrally the [OI] line in the (-570 km s-1 160 km s-1) interval and subtracted the resultant image with a PSF obtained by integrating the line-free range of the spectra (6260 Å - 6270 Å). The resulting image, already slightly elongated in the jet direction, was deconvolved using the lucy procedure (Snyder et al., 1993) in IRAF/STDAS 2. lucy requires a noise estimation and a background subtracted image. The histogram of the image to deconvolve was fitted with a gaussian, thus measuring the sigma and centroid. These were used, respectively, as noise and background estimators. The final resolution (FWHM) for the deconvolved image was 0.24", convergence being obtained after 26 iterations.
Fig. 3 shows the deconvolved image, the jet is clearly visible, it points in the expected direction: position angle and extends to the limit of our field of view were it gets contaminated by a deconvolution edge artifact (the broadening at ). The jet axis points to the embedded primary and originates from an unresolved peak coincident with the system. We deconvolved the integrated cube in the [OI] line (-570 km s-1 -120 km s-1) zone and this peak is still present. The integrated cube at lower velocities (-120 km s-1 160 km s-1) was deconvolved and found unresolved. The origin of this peak can be twofold: due to a continuum underestimation in the subtraction or because the jet is already "super-alfvénic" at our spatial resolution of 230 AU. Our precision in the continuum subtraction precludes the first possibility. The second is supported both from jet theory, where the jet is already super-alfvénic at linear scales of the order of 10 AU (e.g., Ferreira 1997), whose synthetic [OI] jet maps (Cabrit et al., 1999) have a morphology very similar to our observations and, from integral field spectroscopy observations of DG Tau (Lavalley et al., 1997) and long slit spectroscopy of CTTSs (Hirth et al., 1997). It should be pointed that our precision for the center of this peak is 0.1", although our precision in the subtraction is enough to validate the existence of the unresolved peak. This low precision originates from the continuum subtraction - variations of 2% in the continuum will cause shifts of 0.03" in the unresolved component position (the jet however is unaffected ).
We fitted the jet width along the jet axis with a gaussian and found it to be unresolved. Typical jet widths for Taurus CTTSs at similar distances from the central source are AU (Ray et al., 1996) which would be unresolved at the Z CMa distance.
It is worth notice that the microjet probed by our observations has a dynamical time scale (assuming 500 km s-1 from the [OI] line profile) of yr. We associate this material to the 1987 outburst ejecta.
© European Southern Observatory (ESO) 1999
Online publication: June 17, 1999