Use different qcoptions for optimisation and torsion scanner

[howlfwq]

Hi,
Since torsion scanner really takes time and ANI convergence issues on optimisation, I'm wondering if qubekit can use gaussian/psi4 for optimization/hessian calculations, and then use ANI only for single point energy calculations of tosion scan.

[jthorton]

HI @howlfwq, this is something I have been looking into while developing Bespokefit and we would love to add this to QUBEKit as well. For now we are working on fragmentation to help speed up the torsion scans then we can look into this.

[howlfwq]

Hi @jthorton, when I use ANI to only run the torsion scan, I got the error: "geometric.errors.GeomOptNotConvergedError: Optimizer.optimizeGeometry() failed to converge." Can I only calculate the energy rather than optimize the structure in the torsion scan part?

[jthorton]

I got the error: "geometric.errors.GeomOptNotConvergedError: Optimizer.optimizeGeometry() failed to converge."

Unfortunately, this is common when using Geometric and the ani model to perform optimisations and we haven't found a way around it yet.

Can I only calculate the energy rather than optimize the structure in the torsion scan part?

Currently, this is not an option in the default QUBEKit workflow, but with a small amount of scripting, you could do this manually and insert the data into the QUBEKit workflow. How do you plan to generate the geometries for the energy calculations?

[howlfwq]

I was thinking we just use xtb to optimize the geometries generated by the qubekit during the torsion scan and then use ANI to calculate their energies. A JCIM paper has done this with ANI. Here is the DOI:https://pubs.acs.org/doi/abs/10.1021/acs.jcim.9b00439