[lammps-users] Interatomic Potential

Hello!

An interatomic potentials repository is accessible in the link below:

http://lammpstube.com/mdpotentials

This repository provides a source for interatomic potentials, related files, and evaluation tools to help researchers obtain interatomic models and judge their quality and applicability.

good luck!!

Hi
Thanks a lot for sharing. But I check some elements in the periodic table and there are only Lennard-Jones parameters. Is it the case for all elements ? H and He have same parameters ?
Thanks again
Best

Pascal

Pascal,

It would be a big mistake to apply potential parameters without checking what specific system and conditions they were parameterized for. Also classical empirical potentials are based on making some significant approximations which makes them rather specific and non-transferable, so there cannot be just one set of parameters per element. Just look at some of the bio force fields where you have tens of different atom types for carbon atoms depending on their environment. Add to that the fact that for any polar group of atoms you need to overlay the Lennard Jones interactions with Coulomb interactions, which are usually residue or environment specific, you quickly get to the point where a periodic table of elements Form of display is very misleading. To make matters worse, there also is the fact that the parameter sets for different families of force fields are balanced differently, so combining those can incur significant errors.

Bottom line, pure Lennard Jones parameters have limited applications to noble Gases and bulk metals (but for those you have better models), and mix and matching them freely is extremely risky and more likely to cause bogus results than not.

Axel.

You can see how to list and document parameters for MD simulations is done properly by checking out the OpenKIM website at https://openkim.org/
Axel.

Thanks Axel,
This is exactly what I meant, and I was surprised by the announcement.
Thanks again.
Pascal

Hi Axel
Fine with openkim. Looks a very powerful database. With many-body FF.
Best regards
Pascal

Pascal Brault
DR CNRS
GREMI UMR7344
CNRS-Université d’Orléans

Hi Axel
Fine with openkim. Looks a very powerful database. With many-body FF.

many-body potentials and potentials for atomic species make the most sense for the representation in a periodic table-like display. this still needs to be looked at with great care as - for example - interactions of oxides or sulfides often needs special treatment of electrostatics, e.g. though charge equilibration. this is why it is so very important that you have not only the parameters but also inputs and results that characterize the properties of a set of parameters which then allow to a) select the most suitable parameterization for a specific problem (or determine whether a suitable set of parameters is present at all) and b) make certain that the parameters are used correctly by reproducing the recorded properties. OpenKIM goes one step further by (typically) embedding the force kernel itself, so you are literally using the same code to compute the same interactions in all supported MD codes.

as already mentioned for most molecular empirical potentials, the subdivision should not be by element but by atom type and environment (which can be different for different force fields) and that doesn’t include the required assignment of (partial) charges and enforcement of rules/conditions related to those. the correct choice of atom type is crucial since it usually (and crucially) also determines the choice of parameters for bonded interactions which are primarily responsible for local interactions and the local structure, while non-bonded interactions are more responsible for aggregations and interaction with solvent. often molecular potentials are parameterized for a specific solvent, too. in general, (empirical) molecular force fields are complex balancing act that require some degree of understanding of the whole process of parameterizing for a given force field. the major exception to that statement are bio systems, where large databases of standardized templates (residues) exist (e.g. amino acids, nucleic bases, sugars etc.) and the atom typing and parameter assignment can be deferred to a database lookup.
but that again, that process will break down, once you have a non-standard residue in your system that cannot be easily derived from the existing parameters for common modifications (by applying so-called “patches”). then one has to follow the protocol and systematically generate and optimize the missing parameters and specifically the charge distribution following the force field specific rules, which can be quite challenging (or else there would be more people doing this and fewer questions on mailing lists for tools that provide frontend scripts to help with the parameterization procedure, where it can be automated/scripted).

axel.