[fireflies]

hi guys:
i have a 2D euler unstructured(triangle) solver with the cell centered fvm.
there are 2 bc types: farfield and wall. to calculate the flux through the farfield faces, the ghost cell have been used and only used at the farfield.
1. if i fix the variables of ghost cell equal to incomingflow value with no consideration about inflow and outflow, my implicit sovler works fine.
2. but when i implement the nonreflecting boundary condition at farfield(similar to the way in cfl3d, with 4 types, supersonic_in, supersonic_out, subsonic_in, subsonic_out), after some iteration step, negative density and pressure appears, iteration failed.
i have checked the solution before the step of iteration failed, the flowfield looks fine, it seems that the negative density and pressure had sprung up somehow.
my test case is naca0012 foil, ma=0.8, aoa=1.25.
a explicit solver(Runge-Kutta 3) works fine with the both of the two treatments to the farfield. now i am in a big puzzle about the appearance of negative density and pressure, dose it have anything to do with the nonreflecting boundary condition? dose any one have some suggestions？