Hi:

I have a cube with periodic boundary conditions in two dimensions and shrink

wrap conditions in the third (call it z). I am using the NVT ensemble.

correction: you are *not* using the NVT ensemble, you are using the

"fix nvt" integrator.

I am wondering about the volume part of the computation. What is the volume

piece of the shrink wrap dimension? Is it whatever the instantaneous

there is no "volume part" of the computation. fix nvt disregards the

volume completely. please see below.

envelope of the z dimension is? Or is there something else going on? That

is, V is value of the three bounding dimensions are. Whether the dimension

is fixed or not.

your logic is inverted from the one that is employed in LAMMPS. none

of those n?? fixes *enforce* any kind of statistical mechanical

ensemble. it is the other way around, they are a prerequisite for

obtaining the desired ensemble.

fix nve does plain velocity-verlet style time integration AND NOTHING

ELSE. if you have a well equilibrated system, with fully periodic

boundaries and *NO OTHER* manipulations of the system, you will sample

the NVE statistical mechanical ensemble, but only then.

fix nvt adds to this manipulation of the kinetic energy through

coupling it to nose-hoover chains which, under the same conditions as

above, will result in an NVT ensemble, i.e. the same as if you had an

infinitely large NVE system equilibrated to the desired temperature.

fix nph adds manipulation of the volume in response to pressure.

fix nph adds both, manipulation of the kinetic energy and the volume.

similar manipulations can also be obtained through other means (fix

langevin, fix press/berendsen, fix temp/berendsen, fix temp/csvr and

so on) and people argue whether those truly sample the corresponding

statistical mechanical ensembles. some (e.g. langevin and berendsen)

have known and sometimes significant deficiencies, yet how much impact

those have, depends on the particular scenario.

regardless of that, a system with non-periodic boundaries always

represents an *infinite* volume, whether lammps uses fixed or

shrinkwrap (or m) boundaries is merely a question of practicality (no

point in doing bookkeeping on empty space). it has no impact on the

statistical mechanical situation and none of the conventional

statistical mechanical ensembles apply to such simulations.

axel.