How to Perform Hybrid Functional Calculations on La2O3 and CeO2 without ADMM?

[yis...@163.com]

Dear CP2K developers and users,

Hello everyone,

I am currently working on the Pourbaix diagram of lanthanide elements and would like to use CP2K to perform hybrid functional calculations for La₂O₃ and CeO₂. However, since there are no default ADMM basis sets available for lanthanides, the calculations are extremely slow—to an extent that is hardly acceptable.

For systems doped with La and Ce, I managed to proceed by using the auxiliary basis sets of La and Ce as the primary basis sets. Although this approach also slows down the calculation, it is still acceptable. However, when it comes to hybrid functional calculations for pure La₂O₃ and CeO₂, the computation becomes practically infeasible.

What should I do in such a situation? I found the def2-SVP basis set on the website https://www.basissetexchange.org/, and I’ve heard that this basis set should be used together with the SDD pseudopotential. May I ask a beginner-level question: where can I download the SDD pseudopotential? Is it the same as the ECP_POTENTIAL pseudopotential provided in the CP2K folder?

I would greatly appreciate any guidance or suggestions from the community. Thank you all in advance for your help!

[Jürg Hutter]

Hi

For correct calculations you need to have consistent pseudpotentials/basis sets.
For example
LnPP1_POTENTIALS / BASIS_MOLOPT_LnPP1 (I would take those)
or the potentials and basis sets from
https://www.basissetexchange.org/

At least for Ce you can find more possibilities in the (CP2K) literature.

For HFX/ADMM calculations you need corresponding ADMM basis sets. I would suggest to
produce your own. Check the atomic code input examples on how to generate a basis and
follow the advice in the original ADMM paper.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of yis...@163.com <yis...@163.com>
Sent: Wednesday, October 8, 2025 5:24 AM
To: cp2k
Subject: [CP2K:21882] How to Perform Hybrid Functional Calculations on La2O3 and CeO2 without ADMM?

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/4dfbd57d-a432-4d2f-96fc-6696967ab827n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/4dfbd57d-a432-4d2f-96fc-6696967ab827n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[yis...@163.com]

Thank you for your reply. Your previous answers have resolved all my confusion. As a beginner in CP2K, I have one further question: If I want to construct ADMM auxiliary basis sets on my own, are there any specific examples or websites with relevant tutorials that I could refer to? Additionally, could you please briefly explain the main steps for building ADMM auxiliary basis sets, as well as how the quality or success of a self-constructed auxiliary basis set is typically evaluated?

[Jürg Hutter]

Hi

If you are using the LnPP1 PP and Basis you have 26 basis functions per La atom.
The most diffuse Gaussian is 0.03, which will cause problems (CPU time) in the HFX
calculation.
I would generate (using the atom code) a basis for La with 3s 2p 1d (=14 functions) and limit the most diffuse Gaussian to approx. 0.15.
Together with the cFit3 basis for oxygen, you should be able to do the calculations.

best

JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of yis...@163.com <yis...@163.com>

Sent: Thursday, October 16, 2025 3:38 AM
To: cp2k
Subject: Re: [CP2K:21902] How to Perform Hybrid Functional Calculations on La2O3 and CeO2 without ADMM?

To view this discussion visit https://groups.google.com/d/msgid/cp2k/e1d5e266-98a9-4bcb-847b-f41e31a0506en%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/e1d5e266-98a9-4bcb-847b-f41e31a0506en%40googlegroups.com?utm_medium=email&utm_source=footer>.