Dear LAMMPS users,
Do you know an implementation of the solid geometry for DPD simulation
in LAMMPS? What I mean is the following: since the forces are soft, it
is not enough to create a walls made of frozen particles like in MD,
one also need to add bounce-back. And another thing is how to specify
such geometry.
If it doesn't exist, would it be good to have such a package in lammps
(one define geometry and run flow simulation in it)?
Dear LAMMPS users,
Do you know an implementation of the solid geometry for DPD simulation
in LAMMPS? What I mean is the following: since the forces are soft, it
is not enough to create a walls made of frozen particles like in MD,
one also need to add bounce-back. And another thing is how to specify
such geometry.
this topic came up very recently. you can use immobile particles *and*
a reflective wall (or a suitably defined region with fix oneway
instead of the wall))
axel.
A couple ideas.
The LIGGGHTS package, an extension to
LAMMPS for granular systems, reads in triangulated
surfs and runs hybrid particle/surf simulations *and
a lot of other stuff). Those
are for frictional interactions between surfs and particles.
The fix wall/region command can define a fairly
complex boundary (using union or intertesect regions)
to surround particles. Nothing like an arbitrary triangulated
surface however. The fix wall/region command has a few
styles of wall/particle interactions. More could be added.
Adding a granular interaction would be a nice thing to do,
but non-trivial.
Steev
Kirill,
I am curious. Would it not be enough to simply change the coefficients for the interactions between “frozen/wall” particles and DPD particles? If the interaction is sufficiently stiff, particle overlaps are small and the phases (wall vs. free) cannot pass through one another. I mean from what I can tell, the DPD interaction is merely a variant of a generalized granular impact model (with varying powers for overlaps and impact velocities) + a random gaussian term. I mean, for example, what Yiding was asking is taken care of by the definition of the DPD interactions themselves.
eric,
the problem with dpd is that the particles are "soft at the core",
i.e. they can move "through" each other, so you have to find a way to
send those (few) particles that pass through a wall back to where they
belong and a reflective wall or a velocity component flipping region
(fix oneway) are the simplest way to address this issue. ...and
readily available in LAMMPS.
cheers,
axel.
Last addition and I will leave it.
I understand that the DPD interaction is much softer. However, from what I can tell the wall layer of DPD particles is present merely to scatter momentum (and add sufficient friction near the wall), with an additional rigid boundary to keep stuff in the box. The notion of a correct way to accomplish this appears quite arbitrary to me, if that is all that needs to be accomplished. Aside from trying to accomplish what’s been done in prior literature (following some convention), there seems to be a ton of ways to do this. I would presume that there are other ways than simply using a rigid wall that may provide more freedom to lammps users, while remaining no less nor more ad hoc.