# Instability in Lattice Boltzmann Solver

I wrote a Lattice Boltzmann Solver in Rust a little while back using both the BGK approximation and the TRT (two relaxation time) method on the D2Q9 lattice. In both cases I run into major stability problems even in simple examples. Here is a video of the output from the simulation of a left-to-right flow with a cylindrical obstruction (using the bounce-back method): link. This video shows momentum density using lightness as magnitude and color as direction. I can't remember if it was generated with BGK or TRT but I experienced similar failures with both collision operators. I'm using convolution with a [[1, 1, 1], [1, 0, 1], [1, 1, 1]] stencil to implement the streaming step. I tried following the steps described in The Lattice Boltzmann Method: Principles and Practice for getting stability in BGK and TRT simulations, but haven't had any luck. Am I doing something obviously wrong?

• What flow regime are you running, I know that LBM has problems with stability in high speed flows and requires large stencils in those cases
– EMP
Apr 13 '19 at 5:42