search

grimme-lab / xtb

Semiempirical Extended Tight-Binding Program Package
https://xtb-docs.readthedocs.io/
GNU Lesser General Public License v3.0
575 stars 144 forks source link

Manual modification/assignment of GFN-FF topology #250

Open andrewtarzia opened 4 years ago

andrewtarzia commented 4 years ago

Hello all,

I am testing out the new GFN-FF and CREST implementations on some metal-organic systems and it is really great, thank you for the excellent work!

I have some systems in a screening workflow that are quite far from equilibrium structures that I would like to optimise with GFN-FF. However, the automatic topology assignment will often fail. So I was curious if there was the potential to either manually edit the topology that xtb produces OR to provide xtb with a bond topology? Are these features we might expect in the future?

Thank you again, and apologies if I have missed something in the docs that would have answered this question.

Andrew

awvwgk commented 4 years ago

The easiest way to provide a bond topology would be a CT format (sdf or mol), but currently this topology is only used on 2D input to convert them to a 3D structure. One could make this topology available for the 3D input as well, it will miss some details of the actual used topology, but might be useful to guide the topology generation.

An easy to inspect and modify text representation of the complete topology is something we still need to develop.

andrewtarzia commented 4 years ago

Thank you for that. Have you found that the 2D-3D conversion worked for the metal-organic cage systems in the Angewandte paper? I assume they might be too complicated to go from 2D-3D automatically, but would be happy to be wrong.

sespic commented 4 years ago

We did not try the 2D-3D conversion for the metal-organic systems. We took the structures directly from the CSD. Something that I successfully did in the past, was the pre-optimization with GFN2-xTB, if the system size allows it. Here, I am talking about only 20-40 optimization cycles. The outcome is usually good enough for GFN-FF to work.

andrewtarzia commented 4 years ago

Thank you, I may give that a go. Currently using UFF4MOF in GULP to preoptimize the structure. I think the issue is actually with GULP/UFF4MOF. Thank you for your help.

jevandezande commented 7 months ago

Any updates on manually specifying the topology of 3D structures? I'm looking to generate SDF files with r-groups swapped out, and it would be great to perform an initial GFN-FF optimization to help with steric clashes.