Analytic second derivatives for DFT

Geometry optimization and DFT in psi4 is very fast: comparable to ORCA 5, and much faster than NWChem (or at least easier to configure to be fast, i.e., density fitting).

However, calculating the Hessian matrix with DFT using numerical methods is painfully slow. I also cannot find a way to restart a frequency job that stops partway through, so they really waste a lot of computer time.

When are analytic frequency methods going to be implemented in psi4?
And if they aren’t planned, what is the obstacle to their implementation?

Thanks again for all the hard work on psi4!

Analytic frequencies are desired but not planned. We only have so much time to develop Psi4, and the groups who do the most work on that are focusing on other areas. (To give you a smattering: new SCF algorithms, new SCF convergence techniques, CC properties, DLPNO methods, f12 methods…) None of the major contributors are DFT specialists.

If you or somebody else wants to contribute these to Psi4, hop on our Slack channel and ask for guidance.

Well, those features would be even better. I didn’t know those were in the pipeline.

Are the planned features listed on the github?

And I will consider jumping on the Slack. I am an experimentalist but really would like to help develop quantum chemistry software, and psi4 seems like the natural place to start.

There isn’t a comprehensive list of planned features, I’m afraid! Some developers will mention one if it becomes relevant.

The Psi community is fairly beginner-friendly, so feel free.