2025-04-03 JAC/LW check-in

Levels of Theory

• Based on what Peter Eastman recommended I found: hf3c, pbeh3c, b973c, r2scan3c, and wb97x3c

• I think we could get a paper out of comparing these with the methods:

  • that stood out from my literature review: b3lyp-d3bj, b3p86, b3p86-d3bj2b, pbe0-d3bj2b

Paper:

• I picked out 47 unique molecules from our primary dataset and reviewed the literature to get cif files for 9 Li complexes, another Mg, another Zn, and 8 Pd complexes for a total of 66 in our small dataset.

• Compare methods listed above with additional reference ab initio methods with HF, ccds(t) (for small molecules), and scs-mp3?

• Psi4 doesn’t have many basis sets that support Pd, so we will go with def2-TZVP and def2-tzvp-ri (should dramatically speed up? check with low LOT). There is also dzvp

• I think the Grimme 3C methods are supposed to have their own optimized basis sets but I don’t see those in Psi4

• Will compare RMSD with crystal starting structure, and with scs-mp3 if validated against CCSD(T) for small molecules. Also compare bond distances between metal and neighboring atoms.

• Compare energies at Opt and at scs-mp3 LOT

• For CCSD(T) we can “cancel” records in QCA by changing the status

  • LW: Add 10 small organic molecules, one from each class in Pavan’s benchmarking paper

Chodera TM

• Running but difficult to get resources

  • Some workers can’t find nodes due to lack of resources

  • MW >= 1600 Da are having difficulty finding resources. We may need to cut those or find alternative compute pathways.

For PDB CCD

• Waiting for new release to run and get more meaningful errors, moved to Science Backlog until then.

• or just submit with B86/def2-TZVP? I’m thinking wait for the study mentioned above.

  • Let’s ask Chris if he needs this or not

Zenhub:

Tickets:

• Consider closing #699 since it won't handle ferrocenes