The nominally self-preserving mixing layers that originate at the upper and lower lips of a ‘two-dimensional’ jet nozzle meet at a downstream distance x equal to about 6 times the nozzle height h, but self-preservation, this time as a fully-developed jet, is not regained until x = 20h or more. In the intervening region, and possibly in the fully-developed jet, the flow can be thought of as two interacting mixing layers. Dimensional quantities like the mean velocity and Reynolds stresses are of course altered by the interaction, but it may be expected that the dimensionless structure parameters of the turbulence will be less affected.

In interacting turbulent boundary layers in plane duct flow (Dean & Bradshaw 1976) the turbulence structure of each boundary layer is not significantly altered by the interaction, which means that even a fully-developed duct flow can be predicted by a calculation method using empirical data derived solely from isolated boundary layers (Bradshaw, Dean & McEligot 1973). Similar calculations for jets and wakes (Morel & Torda 1973) were less satisfactory, and implied significant structural changes due to the interaction.

The present experiments were designed to investigate the structural changes in a jet in still air. Fluid originating from one mixing layer was permanently marked by heating, and contributions to turbulence properties were measured separately for the instantaneously ‘hot’ and ‘cold’ zones, as done by Dean & Bradshaw and by other workers. Differences between hot-zone structure parameters in the isolated mixing layers and in the interacting region could then be deduced. The results show that near the centre-line the behaviour of the triple velocity products that effect turbulent transport of Reynolds stress is greatly altered by the interaction, the implication being that the large eddies from either shear layer do not ‘time share’ near the plane of symmetry in the simple way they appear to do in the duct. However these changes near the centre line seem not to extend to the maximum-shear region, and calculations using the superposition procedure, with a more refined basic turbulence model than that used by Morel & Torda, are in quite good agreement with experiment. Further improvement to this or any other calculation method would probably require the introduction of a transport equation for the triple product $\overline{uv^2}$ appearing in the transport equation for $\overline{uv}$.