autodE Workflow Terminates When Even one xTB reactant/product Geometry Optimization Fails During Conformational Search

[adragthest]

Hello,

I have noticed that autodE aborts when, during the reactants/products conformational search, a geometry optimisation fails and the energy of the conformer cannot be retrieved.

I took a look on what might have caused the geometry optimisation to fail (since this is not something that happens frequently), and I have found that some reactants/products initial conformer geometries were nonsensical, with atoms overlapping, bonds being too big etc. Obviously xTB has no chance of optimising these geometries starting from so bad initial geometries.

Obviously these can be ignored and the workflow can proceed, but it seems that autodE brings the entire workflow to a halt and terminates abnormally.

Since this behaviour of the program is quite restrictive (stopping a calculation that might have otherwise been completed successfully and have given sensible results), is it possible to maybe ignore the geometry optimisations that have failed, and use the ones that haven't or at least give the user the option of doing that in case he/she wants to proceed with the calculation despite the failures?

Thanks!

[t-young31]

Hi @adragthest – thanks for highlighting this, the option doesn't currently exist but I think this it's a great idea to implement. It's already partially implemented

autodE/autode/conformers/conformer.py

Lines 112 to 114 in 6a6eebe

	except AtomsNotFound:
	logger.error(f"Atoms not found for {self.name} but not critical")
	self.atoms = None

but I reckon we should be consistent with this with all sorts of calculation failures

[adragthest]

@t-young31 Thank you very much for your feedback! Looking forward for the additions!