I am dealing with a simulation for fluid on top of a rigid wall, but with
long-range Coulombic force included. Let's say Z is the direction
perpendicular to lower wall plane, and X,Y are the direction of lower wall
plane. I want the boundary condition in Z is free boundary. But What
boundary should i specify in Lammps? "fixed" or "periodic" ?
If i specify periodic, then if some atoms are lost from top of domain,
they will not re-enter the domain from the lower of domain, because the
lower wall is rigid solid. So, the periodic boundary is far from reality.
this is nonsense. the atoms will happily cross the periodic boundary.
however, it will remain a periodic boundary, if you put a soft wall
into your system so that they cannot cross. the important point is
that atoms will still be considered for computing interactions across
the periodic boundary and the periodicity matters for the calculation
of the long-range electrostatics, regardless of whether atoms can
physically cross or not.
If i specify fixed, it seems more close to reality. However, the
long-range Coulombic force calculation(PPPM/ewald) will have error and
terminate the simulation. It requires the Z periodic.
yes. but you can use MSM instead which doesn't not require all
boundaries to be periodic. it requires, however, careful tuning of
If i use K_space modify to turn off z direction kspace force, then the
simulation will tell error: Bad kspace_modify slab parameter. But if i turn
off the Z direction kspace force, will the simulation be still good?
probably not, but you better find that out properly for yourself by
reading and understanding the physics of what this does. dealing with
poisson solvers has its pitfalls and those can only be avoided, if you
know what you are doing, and that requires reading some papers and
text book stuff.
It seems there is no way to specify Z top as periodic and Z lower as fixed.
that makes no sense at all. you cannot have that kind of setup. what
should the half-periodic thing be periodic to?
the closest to this would be a mirror boundary (which is not
implemented in LAMMPS).
Any one can give some help? Thanks!
most likely the issues are not the technical issues that you worry
about, but a conceptual problem of setting up the right model and
understanding its implications. this has been discussed many times in
the past and those discussions should be archived in the mailing list
archives. please have a look and then catch up on the corresponding