[lammps-users] question on box/relax command

Dear All,

I am doing tensile test on 30x30x4 lattice constants gold sample at 0K. Boundary conditions are p p s.
I use box/relax sample to stretch/compress.

fix 1 all box/relax aniso 8000 8000 NULL.
When the target pressures are of same sign, the system is reaching the
specified pressure at the end of minimization. 
When i use target pressures of opposite sign, say 8000 -8000, the pressure 
after minimization is
 becoming -3.8 in both directions.
Am i correctly using the commands. Is there any other method to 
stretch/compress the sample for the above boundary
 conditions at 0K.
Could anyone help me

I would ask Aidan about this - he has had success using these options
with applied stress. But he’s gone for a couple weeks.


Remember, any time you specify unequal normal stresses, you are specifying a
non-hydrostatic stress state, and so LAMMPS has to add a strain energy
contribution to the energy that it is minimizing. This strain energy depends
on the arbitrary reference state, which is taken to be the starting box
dimensions. The minimization procedure will converge to a stress state that
will equal the one you specified *only if* the final state equals the
reference state. So, you can only get the right answer when you start with
the right answer. How nice. The solution is to allow LAMMPS to reset the
reference state to the current state periodically using the nreset keyword.
Read about that in the fix box/relax doc page. Also, convergence can some
times be a problem, this stuff is not foolproof. So, try to solve easier
problems first, view all results critically, and build up to what you really
want to do.

Finally, I worry a little that you sample is non-periodic in z. I am not
sure what the consequences of this might be.