o-- In a coarse-grained MD simulation which only involves pairwise inter-particle forces, I have fluid flows wrapped by dense “wall” particles as boundary of tunnels.
o-- I use “velocity … 0 0 0” and “fix … setforce 0 0 0” to freeze the “wall” particles.
o-- However, when the simulation box contains only the frozen “wall” particles during time stepping, I still noticed a positive large value for the system pressure on screen output, but not zero as I expected.
Any comment or explanation is appreciated!
Thanks in advance,
The pressure is formally given by P = -dU/dV, so if the wall-wall
interactions make large repulsive contributions to the total potential
energy, it makes sense that P is large positive, regardless of how you
subsequently change the forces. You can get the result you want (and
speed up your simulation) by excluding the wall-wall interactions
using the neigh_modify exclude option.
Thanks for the explanation, and suggestion.
o-- But when I tried using “neigh_modify exclude” for wall–wall interactions, the results of the fluid flows are different than before, as shown in the attached figures.
o-- I am using many-body DPD as the coarse-grained MD in this case, which involves the long-range attraction and short-range repulsion. I have no idea why the usage of “neigh_modify exclude group_wall group_wall” led to such difference. Generally speaking, it appears that the wetting fluid becomes more wetting…
Thanks for any more comment and discussion.
This seems like a legitimate problem, if that’s the only difference in the script. But is this something anyone can actually help you to debug? From what I understand the distribution of LAMMPS only includes ordinary DPD.
I thought you were using a twobody potential, because you said
"pairwise inter-particle forces." With a manybody potential, this
behavior is not at all surprising. The command "neigh_modify exclude
group_wall group_wall" removes all wall-wall entries from the neighbor
list, which then changes the wall-fluid interactions. Given the
nebulous meaning of wall-fluid interactions in a manybody DPD model,
it is very hard to say which of these two simulations is more
physical. This could be determined by measuring thermodynamic
properties such as surface energy. The value of the pressure is
probably not the biggest concern here.
Aidan is correct. Though, I’m not familiar with what people like to call many-body potentials; in reality many-body dpd is just a mean-field description with a hidden variable(density). Looking through several papers, the definition of your local density term in the conservative force is what is corrupting your calculations. Without wall-wall interactions the computed density of wall particles is incorrect (according to these formulae).
To your original question, your best bet is probably using stress/atom appropriately to hopefully remove wall-wall terms from the pressure.
Great tips. Thanks all for providing the insights in this discussion.