A survey for medium redshift (2.2<=z<3.0) optically variable QSOs.

We describe a survey for variable QSOs carried out for 15 years with the UK Schmidt telescope, using IIIaJ plates exposed behind a Schott GG395 filter. Objects brighter than B=21.0 on a reference plate and displaying a peak to peak amplitude of variability larger than B=0.35mag are selected. Plates in U, V, R and I were also obtained. We show that ~20% of all QSOs are missed because of overlapping images, but that ~95% of UVX QSOs (z<2.2) are variable. A smaller fraction of non UVX QSOs are variable; it is demonstrated that this is not a redshift effect, but rather a luminosity effect. We show than ~20% of the low luminosity (M_B_>-25.5) QSOs have an amplitude {DELTA}B>1.0mag, while no high luminosity QSOs have such a large amplitude. We give a list of 107 QSOs with z>=2.20, 71 of them constituting a homogeneous sample of QSOs brighter than B=21.0, in the redshift range 2.2<=z<3.0, having an amplitude of variability larger than {DELTA}B=0.35mag. We show that this sample is probably 65% complete, 15% of the QSOs being lost because they have not (yet?) reached an amplitude of 0.35mag. We compare our sample to two slitless surveys reaching a similar limiting magnitude and find that our estimate of incompleteness is reasonable; however we emphasize that such comparisons are made difficult by the facts that the photometric systems are not identical and that each magnitude scale could be affected by a non negligible zero-point error.