Based on my experience, pressure-based solver performs okay for the typical nozzle supersonic flows. nI do not see any attachment. Could you post some pictures of your mesh? Have you tried a precursor RANS simulation in order to have an approximation of what the result will look like?nFor a pressure-based solver, a time step of 5e-6 seconds seems too big for a LES simulation of a supersonic flow. I have not used density-based solver, even though I suppose that the time steps of both solvers will be of the same order.n