Error while performing DFT+U calculations on FeO

lyonworton commented 7 months ago

Describe the bug

When calculating DFT+U for FeO, the calculation is incorrect when the k-point density is slightly larger。

Expected behavior

No response

To Reproduce

No response

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).
[X] Fix the bug.
[ ] Test the fix.
[ ] Update documentation (if necessary).
[ ] Close the issue and inform the reporter (if applicable).

dyzheng commented 7 months ago

Thanks for your report, can you upload your input files?

dyzheng commented 7 months ago

Set uramping 0.5 will get correct bandgap, you will see a increasing bandgap curve in SCF calculation:

------------------------------------------------------
  Energy         Rydberg                 eV         
------------------------------------------------------
E_bandgap          0.0000277916         0.0003781246
E_bandgap          0.0000400355         0.0005447110
E_bandgap          0.0001210641         0.0016471621
E_bandgap          0.0000831820         0.0011317487
E_bandgap          0.0001351464         0.0018387618
E_bandgap          0.0000873010         0.0011877909
E_bandgap          0.0000483869         0.0006583377
E_bandgap          0.0000702388         0.0009556477
E_bandgap          0.0002124803         0.0028909429
E_bandgap          0.0001420357         0.0019324948
E_bandgap          0.0000410001         0.0005578353
E_bandgap          0.0001580042         0.0021497576
E_bandgap          0.0003137059         0.0042681876
E_bandgap          0.0058235562         0.0792335472
E_bandgap          0.0206244153         0.2806095659
E_bandgap          0.0352565269         0.4796896576
E_bandgap          0.0579382700         0.7882906049
E_bandgap          0.0668592768         0.9096671282
E_bandgap          0.0672814058         0.9154104880
E_bandgap          0.0776505389         1.0564897817
E_bandgap          0.1029670977         1.4009392355
E_bandgap          0.1025543748         1.3953238526
E_bandgap          0.1026821734         1.3970626410
E_bandgap          0.1110515606         1.5109339962
E_bandgap          0.1341241462         1.8248526282
E_bandgap          0.1336153500         1.8179301003
E_bandgap          0.1352249377         1.8398296650
E_bandgap          0.1414632697         1.9247065261
E_bandgap          0.1595030580         2.1701504367
E_bandgap          0.1596140032         2.1716599240
E_bandgap          0.1605181973         2.1839621156
E_bandgap          0.1618843831         2.2025500272
E_bandgap          0.1645047282         2.2382016513
E_bandgap          0.1664677312         2.2649096769
E_bandgap          0.1663291243         2.2630238345
E_bandgap          0.1658992910         2.2571756518
E_bandgap          0.1658240135         2.2561514487
E_bandgap          0.1656869293         2.2542863221
E_bandgap          0.1656742560         2.2541138929
E_bandgap          0.1656584099         2.2538982956
E_bandgap          0.1656617467         2.2539436952
E_bandgap          0.1656601457         2.2539219131

pxlxingliang commented 7 months ago

I have run FeO with symmetry=1 and 0. And for symmetry=0, the band gap is 0 eV, while symemtry=1, the band gap is about 1.7 eV(U=4eV).

feo.zip

WHUweiqingzhou commented 7 months ago

@maki49 do you have any idea about the reason why symmetry will affect band gap in this case ?

maki49 commented 7 months ago

I did the following tests:

symmetry=1 but comment out the charge density symmetrization: E_bandgap=1.9 eV
symmetry=1, init_chg=file reading charge density and onsite.dm from symmetry=0: E_bandgap=1.7 eV
symmetry=1, init_chg=file reading charge density and onsite.dm from symmetry=0, and comment out the charge density symmetrization: E_bandgap=1.9 eV

So there seems nothing to do with charge density, all the difference results from reducing k-points. Maybe there's some bug with DFT+U when there are too many k-points?

dyzheng commented 7 months ago

symmetry =1 will lead to a rotation of occupation matrix (onsite.dm) with symmetry = 0 or -1, which should be rotate back when symmetry =1. To solve this problem, I'll ask for your help @maki49 later.

mohanchen commented 1 month ago

Any progress for standard systems like FeO?

deepmodeling / abacus-develop