Optimization and Hessian hybrid procedure

[jthorton]

Is your feature request related to a problem? Please describe.

Currently, it is quite awkward to generate Hessain calculations on an Optimizationdataset after it has completed, this involves periodically checking each task has finished and then adding the new calculation on the final molecule.

Describe the solution you'd like

It would be great to have a procedure that could automatically generate an associated base dataset (with the same name as the optimisation dataset?) that contained the hessian calculations without manual intervention.

Describe alternatives you've considered

We could alternatively have a bot per dataset to periodically check the dataset records and submit the new calculations when ready.

Additional context

[dgasmith]

I wonder if this is popular enough that we should encode it into a service?

@mattwelborn?

[mattwelborn]

This feels to me like a good candidate for pipelines?

[dgasmith]

It is, but pipelines will never feel as seamless as an object like torsiondrive. Optimization + Something seems popular enough to be worth considering and can be built into things like conformer search where you often want to do low level optimization + high level energy.

[mattwelborn]

Do you envision an OptimizationPlusSinglePointResult? an OptimizationPlusSinglePointDataset? both?

[dgasmith]

I think the short here is that we need to implement this at the pipeline layer. We will get too many of these custom objects otherwise.

Timeline is ~3 months for a beta, but that is pretty open depending on what else comes up in the meantime.

[bennybp]

I don't have any immediate plans for a pipeline/workflow feature, although I have considered a Optimization + Singlepoint calculation as a simple workflow that covers most use cases.

Superseded by #748