Calculation of centroid distance

GROMACS version:2020.5 or 2021
GROMACS modification: Yes/No

It is also the frame 6027, but the centroid distance calculated by GMX distance is 1.633 nm; When running US, in the " npt6027_pullx. xvg" file, the centroid distance has become 0.004 nm at the beginning.

I don’t know what happened in the middle, which led to this situation? Is it because of the problem of periodic boundary, or is there something to pay attention to ?

You need to specify an appropriate pull-group*-pbcatom for each group, because apparently the geometry of the groups is such that the default of using the numerical middle of the atom numbers lies far away from the geometric center of the molecules. Hence, you get different distances.

As you said, I modified the mdp file and recalculated the distance. Unfortunately, this does not change the calculation results.