Thank you for the reply. I understand your point that NPT is used for
relaxing a structure so as to study bulk properties, and, surface
effects is undesirable. Therefore there is no provision for PBCs with
an NPT ensemble.
That said, I have two queries and the second one motivates why I
(stupidly?) want to use NPT without PBCs:
1) Getting back to the shear example with lammps, it does a NVE
ensemble simulation for equilibration prior to the shear simulation.
This leads to a unrealistically high virial contribution to pressure
in the equilibrated state. I feel that a NPT simulation (with PBCs)
which allows the volume to change and lets atoms occupy positions
within the fcc lattice that are consistent with the empirical
potential used is more appropriate. Is this point of view right?
2) Many researchers seem to use lammps for simulation at the
nanoscale. As the billion atom simulations with LJ potential benchmark
indicates, many nanoscale devices can be simulated using lammps. All
of these have the surface atoms contributing substantially to the
property of the device as the atoms in the bulk. For such simulations
again NPT without PBCs is desirable. I have experience implementing
the Berendsen's barostat/thermostat which works without PBCs.
Therefore the second pair of questions are:
(i) Is it unphysical to want to do NPT without PBCs for such systems?
(ii) Does the implementation of the Nose-Hoover thermostat need
using PBCs with NPT?
lammps has been an eye-opener as far as generality of input file is
concerned. With the meager insight I have had till now into the
documentation, it seems possible to implement a Berendsen's thermostat
within the input files by rescaling all the velocities of the atoms
every time step by the appropriate factor. I am however not sure that
the same can be done to box sizes so as to implement a Berendsen's
barostat by a clever input file. Has anyone done this?
Best Regards and thank you once again.