Problems to read ECP

Dear users,

I want to run a cc2 calculation with ECP for Ag.

I´ve used the def2-svp basis sets which includes the ECP for Ag.

In the output, I´ve got the following information about the basis sets:

Name: DEF2-SVP
Keyword: BASIS
atoms 1, 3-6, 10, 13, 15-17, 20 entry C          line    89 file /home/ancborin/Programs/psi4conda/share/psi4/basis/def2-svp.gbs
atoms 2, 7-9, 12, 14            entry N          line   109 file /home/ancborin/Programs/psi4conda/share/psi4/basis/def2-svp.gbs
atoms 11                        entry AG         line  1482 file /home/ancborin/Programs/psi4conda/share/psi4/basis/def2-svp.gbs
atoms 18-19                     entry O          line   129 file /home/ancborin/Programs/psi4conda/share/psi4/basis/def2-svp.gbs
atoms 21-31                     entry H          line    14 file /home/ancborin/Programs/psi4conda/share/psi4/basis/def2-svp.gbs

That is, there is no mention to the ECP for Ag. The final energies is also wrong.

As it works fine for calculations without ECP, I believe the problem is while reading the basis set.

In the input, I´m using:

set {
scf_type cd
cc_type cd
basis def2-svp
freeze_core true
roots_per_irrep [5]
cholesky_tolerance 1e-4

I will be thankful to receive some input file with instructions on how to use ECP.

Best whishes

Hi Antonio,

First, I want to point out that ECPs are a new feature in psi4 and not yet considered stable, we are still finding and fixing bugs in the implementation. The upcoming 1.2 release will include new ECP features but with a prominent “BETA” tag. And use of ECPs will be documented/advertised more thoroughly in a later release when we have a more stable and more thoroughly tested implementation.

Now on to your issue. I think your problem may be related to this recently posted issue with ECPs and frozen core. You can look over the details of that bug to determine if it is the same issue you are experiencing. If it is not I encourage you to open an issue on GitHub with an example input and expected results, so that the development team can be made aware.

If it is the same bug you are dealing with: The same contributor that noticed the issue has submitted a patch which is currently under review. You can track that PR and once it has passed review and is merged you can rebuild psi4 from source to obtain a version with the patch applied. However I can’t make any promises that this is the only issue with ECPs that you will encounter, and I would suggest that you check results with an external source when using them for the immediate future.

Getting our ECP implementation “production ready” is up there on the priority list, as many users would like that feature to be in a more stable state. I encourage you to keep using it and open issues for any errors you notice so we can track down bugs.

Further to that, you should be able to manually override the number of frozen but explicit electron pairs in the whole system (N) by using set num_frozen_docc N.

For your system, with the def2-svp basis set you want 1 frozen pair for every C, N, O atom, 0 frozen pairs for every H, and as def2-svp replaces 28 electrons in Ag with the ECP (1s^2, 2s^2, 2p^6, 3s^2, 3p^6, 3d^6) you need to add another 4 pairs (for 4s^2 and 4p^6) to get the full [Kr] shell frozen.

This assumes that you’re using a fairly recent version of Psi4 which can read ECP’s…


Thank you, tetrahydrofuran and amjames, very much for your suggestions.

I can wait until you finish reviewing the patch, and try again.

Meanwhile, I will try to use all electron basis sets.

Congratulations for the excelent work you all have done!