I am trying to implement a multiphase cahn hilliard code in libmesh. The code kind of works in single thread mode. However, when I run the code in parallel, I get a discontinuity in the solution roughly corresponding to the partitioning. Previously, when this has occured in my libmesh codes, this has been due to the use of solution instead of current/old_local_solution. I am not using any of these. Is there any other reason that this could be happening? I can post the code if it will help.