[CP2K-user] [CP2K:22457] Suggestion: Improving Spin-Polarized SCF Convergence for Doublet Systems in CP2K

'Jürg Hutter' via cp2k cp2k at googlegroups.com
Fri Nov 6 12:25:53 UTC 2026


Hi
I created an issue on github/cp2k for your request. It might get more attention from developers there.
https://github.com/cp2k/cp2k/issues/6473
regards
JH

________________________________________
From: cp2k at googlegroups.com <cp2k at googlegroups.com> on behalf of Zhen Wu <nangoaaaa at gmail.com>
Sent: Thursday, November 5, 2026 7:53 AM
To: cp2k
Subject: [CP2K:22456] Suggestion: Improving Spin-Polarized SCF Convergence for Doublet Systems in CP2K

Dear CP2K Developers,

I would like to suggest improving CP2K's treatment of spin-polarized calculations for doublet systems, particularly for surface adsorption models such as OH adsorbed on ZnO.

For systems with a relatively small band gap under the PBE functional, I have encountered serious convergence and consistency issues when performing spin-polarized calculations for OH-adsorbed ZnO surfaces.

  1.  Using the OT method: When I apply the standard OT settings that work well for bulk ZnO, the calculation can produce a normal band gap in the α-spin channel, while the β-spin channel becomes metallic. In some cases, the calculated β-spin LUMO–HOMO gap even becomes negative.

  2.  Using the diagonalization method: I have tried different SCF mixing schemes, but none of them can achieve convergence for these doublet spin-polarized systems. Eventually, I can only obtain SCF convergence by enabling smearing.

Beyond the additional computational cost, this creates a serious problem for maintaining consistent computational settings across surface adsorption models. Many of my previously calculated surface structures may need to be recalculated simply because the same calculation settings cannot be applied consistently.

There is also a further problem when I attempt to use the converged PBE results as initial wavefunctions for subsequent hybrid-functional calculations:

  1.  Hybrid functional + ADMM + OT: Because the number of virtual orbitals is inconsistent, the PBE wavefunction obtained with smearing cannot be directly used as the initial guess for this calculation.

  2.  Hybrid functional + ADMM + diagonalization: Keeping the same number of virtual orbitals as in the PBE calculation allows the calculation to run. However, the original problem remains: the SCF mixing schemes still fail to converge for the doublet spin-polarized system.

  3.  Hybrid functional + ADMM + diagonalization + smearing: I have not tested this combination yet. However, I am concerned about whether it can reliably preserve the physically meaningful spin-polarized electronic distribution required for subsequent electronic-structure analysis. Such calculations are particularly important because the purpose of performing hybrid-functional calculations is often to obtain more reliable electronic structures.

Therefore, I sincerely hope the development team will give this issue serious consideration. It would be extremely valuable to improve the robustness of SCF convergence for open-shell doublet systems with spin polarization, especially for semiconductor surface models with small band gaps. Ideally, users should be able to obtain reliable spin-polarized solutions without having to rely on smearing as the only viable approach, while retaining compatibility with subsequent hybrid-functional calculations.

Thank you very much for your time and for your continued development of CP2K.

Best regards,

--
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+unsubscribe at googlegroups.com<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/3620d4c3-3233-4d8b-ab2f-6b7113095df9n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/3620d4c3-3233-4d8b-ab2f-6b7113095df9n%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB07590877ABFF4D9AA6F3A7009F762%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.


More information about the CP2K-user mailing list