lj potential/polydisperse-colloid/epsilon

Dear All,

I know you have been bothered by this question many times by the newbies (if possible, don’t get angry, I apologies in advance).

I have a polydisperse colloidal system (spherical particles with different sizes ranging from 50-500 nm). for the PAIR_STYLE LJ/CUT , I have been trying to get the pair style parameters in REAL UNITS. My biggest confusion is regarding WELL_DEPTH definition(EPSILON)??

pair_style lj/cut 2.5

pair_coeff Part1 Part2 epsilon sigma cutoff
pair_coeff 1 1 * 2000 2000.5 (real units so sigma and cutoff in Angstrom).
pair_coeff 2 2 * 1000 1000.5

pair_coeff 1 2 ** 1500 1500.5 (this becomes even more confusing)

I have been through a many similar simulation literature but everywhere I have found that people just leave description of epsilon with whatever SIGMA to “epsilon is well depth for lj particle interaction strength, deeper the minima stronger the interaction(kind of general description)”.

I have went through other software manuals and method but have not been able to get things right.

If possible could anyone please help me out here with correct literature/method/example??

but if anyone feels offended by such a question, please ignore and accept my apology for posting(not trained in physics just a biologist trying to adapt).

Best regards

A few general comments.

Your parameter choices are very inconsistent: You set a cutoff of 2.5 angstrom (which is extremely short for real units), but have particles with a much larger sigma. That makes no sense. a typical cutoff in real units would be between 8 and 14 angstrom. a typical epsilon between 0.02 and 0.5 kcal/mole

If you are working on colloidal particles, you cannot use the lj/cut pair style, as setting the sigma to a large value will make the interaction very “soft”. You should check out the colloid pair style in the COLLOID package instead, which is a way to handle lennard-jones type atomic interactions to large agglomerate objects.

If you are struggling with grasping the physics, then you should recruit somebody (preferably local) as a collaborator that can train/advise you. Trying to figure this out on your own is likely to lead to either frustration or bogus results or both. A mailing list is not a suitable way to obtain that information, since this is a matter of obtaining a suitable fundamental understanding of the theory and not just advise to get a few keywords right so that your simulation will run.