Mass conservation


I am having a weird issue with convergence – I am dealing with 2D channel flow at large Reynolds number (10000) and following are the details of my run (everything in lattice units):

channel width = 1020
channel length = 3200
viscosity = 0.0018
channel centerline velocity = 0.075

I am adding a perturbation to make the flow unstable, and that does not cause a huge rise in the Mach number. I apply periodic conditions along the channel and the extrapolation scheme of Chen & Martinez on the channel walls. The flow never converges to a statistically steady state; and when I look at the x-component of velocity at the centerline it is continuously decreasing. If I plot \Sigma f_i in the whole domain, it is continuously increasing – indicating mass is not conserved. The code is MPI parallelized.

I have checked for mass conservation at small Reynolds numbers (~10 – 40) and it works well, but has issues at large Re.

Could some one please advise me on the diagnosis?