Support for additional MD methods (e.g. MTK NPT)

[Andrew-S-Rosen]

Hi all! I am opening a Discussion post to suggest some features for torch_sim. One of those is support for additional MD classes, in particular the MTK-based NPT ensemble in ASE. This is now the de facto standard in ASE for NPT simulations and would be great to have in torch_sim as well.

[abhijeetgangan]

Yes, there are plans to extend the NPT to anisotropic cell fluctuations. What are other MD methods you would would like to be implemented?

We just need to be a bit careful to implement them (For eg) - https://gitlab.com/gromacs/gromacs/-/issues/5072.

[Andrew-S-Rosen]

For me, the main one is NPT with anisotropic cell fluctuations. So, it's great to hear! I will think of there are any other critical ones!

[abhijeetgangan]

These are tracked in #301 and #300. We also have other integrations schemes (isotropic) for NPT.

[thomasloux]

the current NPT Nosé Hoover should correspond to MTK NPT if I'm not mistaken, although as previously said only in isotropic version

[abhijeetgangan]

I think the request is for an equivalent of Full MTK NPT as in https://ase-lib.org/_modules/ase/md/nose_hoover_chain.html#MTKNPT.