No Conformers crashing about autode

[kjelljorner]

Well you can get the bond orders with any QC program really... Probably more reliable than covalent radii cutoff anyway.

[t-young31]

Hi @eneas77

Thanks for raising this, as it's a really interesting case.. The problem is that the optimised structure of the N2O3 product contains a very long N-N 'bond', which is not added to the molecular graph of the structure. This means when autodE prunes conformers to remove any with different connectivity (from that given) then no conformers are left and the calculation can't continue. This is going to require a little more thought to fix...

---

Note I found this by setting $AUTODE_LOG_LEVEL to INFO (e.g. add export AUTODE_LOG_LEVEL=INFO to ~/.bashrc) and that in the log the connectivity prune left no conformers then interrogating the molecular graph with:

>>> import autode as ade
>>> mol = ade.Molecule('p1_N2O3.xyz')
>>> mol.graph
MolecularGraph(|E| = 3, |V| = 5)
>>> mol.graph.edges
EdgeView([(0, 1), (1, 2), (3, 4)])
>>> mol.atomic_symbols
['O', 'N', 'O', 'N', 'O']
>>> # Bonds: O-N, N-O, N-O

p1_N2O3.xyz.zip

[eneas77]

Hi @t-young31

Thansk again for the in insights... I suppose that outsourcing the optimization in pure NWChem (no AutodE) and then reinvoking AutodE from the optimized reactans and products won´t circumvent the problem, as the molecular graph is also essential for TS path guess and structure build up... Is that right?

Thanks for your efforts!

E77

[t-young31]

Yep - absolutely right. I have had some thoughts on fixing these intermediate cases but am struggling on finding something general. If you (or anyone else reading this) has any good ideas then I'd be all ears!

[kjelljorner]

What I have done in the past is to use bond order cutoffs to define the graph rather than covalent radii + distances. A quick xtb calculation can bed used for that.

[t-young31]

Thanks for the thoughts – I'm slightly loathed to require a dependency on XTB 😕. I can see this being an interesting research project...