The PBE0 calculation can't converge in systems with a single C atom

[xuan112358]

Describe the bug

When I tried to run SCF in a system consisting of only one C atom with PBE0, I found that I could not get convergent results. I have tried mixing_restart+mixing_dmr, exx_pca_threshold, etc, as 0, and exx_separate_loop as True/False, but none of these worked.

Expected behavior

No response

To Reproduce

Here is an example with exx_separate_loop as False. PBE0_C_github.zip

Environment

No response

Additional Context

No response

Task list for Issue attackers (only for developers)

• [ ] Verify the issue is not a duplicate.
• [ ] Describe the bug.
• [ ] Steps to reproduce.
• [ ] Expected behavior.
• [ ] Error message.
• [ ] Environment details.
• [ ] Additional context.
• [ ] Assign a priority level (low, medium, high, urgent).
• [ ] Assign the issue to a team member.
• [ ] Label the issue with relevant tags.
• [ ] Identify possible related issues.
• [ ] Create a unit test or automated test to reproduce the bug (if applicable).
• [ ] Fix the bug.
• [ ] Test the fix.
• [ ] Update documentation (if necessary).
• [ ] Close the issue and inform the reporter (if applicable).

[QuantumMisaka]

@xuan112358 single C atoms should be a radical, you should use nspin=2 to describe its electronic structure @PeizeLin @maki49 But is it truly a problem ? I have some test which cannot lead to scf_thr 1e-6 convergence (although the energy difference seems to be converged)

test-EXX.tar.gz

it seems EXX update for spin polarized system have some convergence problem ( as it in my FeCx systems?)

@WHUweiqingzhou any comments ?

[kirk0830]

@xuan112358 single C atoms should be a radical, you should use nspin=2 to describe its electronic structure @PeizeLin @maki49 But is it truly a problem ? I have some test which cannot lead to scf_thr 1e-6 convergence (although the energy difference seems to be converged)

test-EXX.tar.gz

it seems EXX update for spin polarized system have some convergence problem ( as it in my FeCx systems?)

@WHUweiqingzhou any comments ?

@QuantumMisaka it is not simply radical, @PeizeLin used to inform me of a keyword "ocp_set" https://abacus.deepmodeling.com/en/latest/advanced/input_files/input-main.html#ocp-set, it can help to get more reasonable result. On the other hand for single atom calculation, the mixing parameter choice, as far as I remember, is a little bit tricky.

[QuantumMisaka]

Interesting @xuan112358 FYI, C dimer which have 1.6 bond length can converge in HSE calculation

[QuantumMisaka]

@xuan112358 In my calculation, single C atom can be converged by HSE. This is my attachments C-HSE.tar.gz

but PBE0 seems to have problem. refer to https://github.com/deepmodeling/abacus-develop/issues/5036

[QuantumMisaka]

In this example above, HSE do not converge to scf_thr 1e-7, but will stuck around. However, the EDIFF can reach 1e-7 eV convergence.