About 'pair'

[Damienfg]

Hi,
I wrote a HOOMD script to simulate the behavior of large spheres (colloids) in a bath of smaller spheres (solvent). I used the DPD framework along with a Lennard-Jones potential as follows:

• The interaction between solvent particles is governed by DPD, as is the interaction between colloids and solvent particles.

• The interaction between colloids is governed by the LJ potential.

When I specified the DPD interactions like this:

dpd.params[('colloid',` 'solvent')] = dict(A=25.0, gamma=4.5)  # Interaction colloïde-solvant
dpd.r_cut[('colloid', 'solvent')] = 1.0
dpd.params[('solvent', 'solvent')] = dict(A=25.0, gamma=4.5)  # Interaction solvant-solvant
dpd.r_cut[('solvent', 'solvent')] = 0.5

I encountered the following error upon execution:
hoomd.error.IncompleteSpecificationError: For <class 'hoomd.md.pair.pair.DPD'> in TypeParameter r_cut for key ('C2', 'C2') value is required

I resolved this issue by including in my script the colloid-colloid interaction in DPD but setting A=0:
dpd.params[('colloid', 'colloid')] = dict(A=0.0, gamma=4.5)

It seems that when using 'pair' —regardless of the potential (here, dpd = hoomd.md.pair.DPD(nlist=hoomd.md.nlist.Cell(buffer=0.4), kT=1.0)) — HOOMD applies the potential to all particle pairs in the system. Defining dpd.params[('colloid', 'solvent')] and dpd.params[('solvent', 'solvent')] does not de facto exclude dpd.params[('colloid', 'colloid')].

Is this correct? Is there a way to specify a potential for only selected pairs?

Thanks, Damien

PS: This is important to me also because my system contains only six colloidal particles (C1, C2, …, C6), with a single instance of each. I need to know how to specify the system so that interactions like C1-C1, C2-C2, etc., are not included, as they do not exist given that each particle appears only once.

[joaander]

Yes, you must set parameters for all (unordered) pairs of types, including ['colloid', 'colloid']. If you do not set them, then pair energy contains undefined terms:

$$U_\mathrm{pair,total} = \frac{1}{2} \sum_{i=0}^\mathrm{N_particles-1} \sum_{j \ne i, (i,j) \notin \mathrm{exclusions}} U_\mathrm{pair}(r_{ij})$$

HOOMD-blue doesn't know that you intend to place only a single particle of each of those types. What if someone later executes the script with a different input file that contains more than one? Would you prefer undefined behavior where HOOMD-blue computes forces using whatever unknown parameters happen to be at that memory location instead of getting an error message?

In your case, you need to set parameters between all colloid types and the solvent:

dpd.params[(('C1', 'C2', 'C3', 'C5', 'C6'), 'solvent')] = dict(A=25.0, gamma=4.5)
dpd.r_cut[(('C1', 'C2', 'C3', 'C5', 'C6'), 'solvent')] = 0.5

and set r_cut=0 to achieve your goal of no same-type colloid interactions:

for t in ('C1', 'C2', 'C3', 'C5', 'C6'):
    dpd.params[(t, t)] = dict(A=0, gamma=1)
    dpd.r_cut[(t, t)] = 0

Yes, you still need to set A, and gamma. HOOMD-blue must assume that you might later change r_cut for these interactions.

Of course, you will also need to set the full matrix of Cn-Cm interactions as well.