Kelvin-Helmholtz instability in a confined geometry

Growth of kelvin-Helmholtz instabilities received many attention in the case of wakes or shear layers with thickness increasing in the downstream direction. In contrast with these ``unstationary'' situations, few works investigate stationary situations in which a Kelvin Helmholtz raw grows and saturates inside an imposed geometry. One of the sole exception is an experiment developed by M. Rabaud \& Y. Couder where vortices develop between concentric rotating disks. We developed an experiment in a different geometry: a recirculating belt is running at the free surface of a long rectangular tank of larger width. In some conditions a regular rectilinear raw of vortices develop between the flow dragged by the belt and the flow recirculating beside the belt. When the belt is on the central axis of the tank, with free surface on each side, recirculation flows on each side. In this case two regular raws can develop. Vortices of these two raws can interact and a 180$^{\circ}$ phase shift between them is observed. Properties of these vortices have been investigated and we tried to relate it to available models. Parallels with jet instability can be made. We also have explored the occurrence condition of the recirculation: at low reynolds number it can occur via the bottom of the tank.