Hybrid mapping in VLBI and spurious symmetrization

The standard procedure used in hybrid mapping of VLBI data is examined. That this procedure can converge on wrong solutions has been pointed in the past by many authors. Walker (1996) indicated the bias in the resultant data due to the use of a point source as starting model. Baath (1989) suggested the use of the original data set in each iteration of self calibration, and not the use of the corrected one created by the previous iteration. Linfield (1986) analyzed the role of the u - v coverage and lack of intermediate spacings. However, the full procedure needed to avoid false features is at present not clear and only experience with imaging helps the user to avoid them. In this paper we demonstrate that the problem is connected with the non-linear nature of self calibration which leaves initial wrong assumptions frozen in the final solution. We demonstrate how large closure phase values linearize the process and we show a simple modification of the standard procedure which helps to avoid creation of false features in the imaging.