search

CompPhysVienna / n2p2

n2p2 - A Neural Network Potential Package
https://compphysvienna.github.io/n2p2/
GNU General Public License v3.0
214 stars 81 forks source link

Symmetry functions with different cutoffs #153

Closed Matheus-Novoa closed 2 years ago

Matheus-Novoa commented 2 years ago

I'm testing the Ethylbenzene_SCAN potential, present in examples/potentials folder. I notice that, in the symmetry functions section, each symmetry function line has a different cutoff value.

My question is which cutoff value i have to use in "pair_coeff" in lammps.

philippmisof commented 2 years ago

As it is described here: https://compphysvienna.github.io/n2p2/interfaces/pair_nnp.html#description

The maximum cutoff radius of all symmetry functions is the only argument of the pair_coeff command which should be invoked with asterisk wild-cards only:

The pair_coeff command basically tells LAMMPS the cutoff radius for the requested neighbor list which is constructed inside LAMMPS. Since the same neighborlist is used for all symmetry function calculations you therefore need to choose the largest cutoff.

Matheus-Novoa commented 2 years ago

I inserted the largest value of symnetry functions cutoff and worked.

I tested the etilbenzene_SCAN potential example with the lammps interface, but i got extrapolation warnings.

I guess this problem is out of scope of this issue, then i will open another one.

Thanks for your help.

philippmisof commented 2 years ago

The extrapolation warning means that the potential is used for configurations which are beyond the range of the data set used for training. Thus you can't use the potential for your current MD run since the potential wasn't trained for your use case (e.g. not for the temperature in your MD). Unfortunately I wasn't involved in the development of this potential so I don't know any more details. But I agree, if you want to discuss this topic further it would be best to open a new issue for that.

You're welcome