I am not sure where, but confirmed there is a problem on compute cna/atom command. That is, when using MPI, the output value of CNA is not always the corresponding value the atom should be. The problem do not occur on the serial version of LAMMPS.
For example, I use the command:
dump 1 all custom 100 cna.test id x y z c_myCna
Unfortunately, the output c_myCna value is not always the atom (x, y, z) should be (partly right), although the distribution of CNA is correct.
Can anyone give me any suggestions on how to solve the problem?