search

ACEsuit / mace

MACE - Fast and accurate machine learning interatomic potentials with higher order equivariant message passing.
Other
523 stars 195 forks source link

Is it possible to develop ReaxFF force fields in LAMMPS-MACE? #641

Closed Matt-A-Thompson closed 2 weeks ago

Matt-A-Thompson commented 2 weeks ago

I am entirely new to ReaxFF, and wanted to know if this is possible (I couldn't find anything in the documentation).

ilyes319 commented 2 weeks ago

I am not sure I understand the question. ReaxFF and MACE are two different packages. What would want to do exactly?

Matt-A-Thompson commented 2 weeks ago

Effectively, I have a system I'd like to develop a ReaxFF force field for, which requires re-parameterisation of an existing model (i.e. a standard MACE interatomic potential (one which does not allow bond breaking, etc.)). I know that LAMMPS can run ReaxFF, and so is it possible to have MACE work with a FF designed for ReaxFF? I have DFT training data on the system, but an actual workflow is something I'm uncertain of. I know that Skylaris et al (https://chemrxiv.org/engage/chemrxiv/article-details/66fd266d12ff75c3a1278cec) have written about this (sort of, the paper compares existing ReaxFF to new MACE-ML models), but I couldn't find any discussion on whether MACE can accept and train ReaxFF force fields.

Matt-A-Thompson commented 2 weeks ago

Apologies - very new to this!

ilyes319 commented 2 weeks ago

MACE is reactive so if that is what you want to do, you can directly use MACE. If you have training data, you can train a MACE (or finetune one of our foundation model like mace-off) and run your simulations in lammps.

Have a look at the tutorials to see how this works.

Matt-A-Thompson commented 2 weeks ago

Thank you so much for clarifying, and apologies for the confusion - I'll go ahead and close this!

gabor1 commented 2 weeks ago

Just to clarify further, mace and reaxff are alternatives to each other, you don't use mace to fit reaxff and vice versa. both can cope with reactivity.

Matt-A-Thompson commented 1 week ago

Thank you for this!