Sorry, I did not pay attention to the email address.

The stress measurement is based on what we have on LAMMPS manual:

compute peratom bulk stress/atom

variable meanpress atom -(c_peratom[1]+c_peratom[2]+c_peratom[3])/3

Also I solved the problem, and to not leave the thread without any responses

I have attached 2 input files, one with the error and the other one without

error:

with error:

compute myCN bulk centro/atom 12

compute myKE bulk ke/atom

compute myPE bulk pe/atom

compute myCOM bulk com

compute peratom bulk stress/atom

compute vz bulk property/atom vz

WO error:

compute myCN bulk centro/atom 12

compute myKE bulk ke/atom

compute myPE bulk pe/atom

compute myCOM bulk com

compute peratom bulk stress/atom

compute vz bulk property/atom vz

It is quite interesting how changing the group option in compute commands

can change the structure (internal properties of simulation). Changes that

alter how the simulation behaves. I am not an expert, but I would say it is

some kind of bug.

there is no bug. what you get is what you are asking for. if you look

at the trajectory files with a text editor/viewer, you will see that

the coordinates are *exactly* the same in both cases. only that the

additional computed properties are only computed for a subset of the

atoms and set to zero in the "error" case. however, that is exactly

the expected behavior when you do a compute on a subgroup of atoms.

since this difference is not a small one, but between zero and

non-zero, one has to assume that the reason for seeing the undesired

behavior is not in the computation of those properties itself, but

rather in how this group of atom is defined.

looking at the group definitions it doesn't take much to recognize

that the "bulk" group, i.e. the one used in the "error" case is based

on a region defined with "box" units, whereas the rest is defined in

"lattice" units, which results in a scaling factor of almost 4, which

is commensurate with the size of the block that is visible in the

visualization.

in short, the bug is in your input (or in your lack of properly

checking it) and not in LAMMPS.

axel.