Menu

My DFT+U scf calculation can not get converged, can anyone please help me to have a look?

Elk Users
Zhiwei Li
2021-11-01
2022-11-12
  • Zhiwei Li

    Zhiwei Li - 2021-11-01

    As described in the title, I was not able to get my DFT+U calculation to converge.
    When I remove the DFT+U part, it can converge with elk-default settings.
    When I do AFM calculations, it also converges with default settings.
    Following is my elk.in for nonmagnetic settings. Can anyone have a look and help me with the settings?

    tasks
    0
    10
    22

    xctype
    3

    mixtype
    3

    ! DFT+U block
    dft+u
    1 4 : dftu, inpdftu
    1 2 3.0 : is, l, lambda

    ! conduction local-orbitals required
    lorbcnd
    .true.

    ngridk
    4 4 4

    nempty
    4

    msmooth
    2

    lmaxi
    2

    maxscl
    200

    sppath
    './'

    nwrite
    10

    avec
    4.901665000 8.489932822 0.000000000
    9.803330000 0.000000000 0.000000000
    0.000000000 0.000000000 26.80387500

    atoms
    3 : nspecies
    'Co.in' : spfname
    8 : natoms; atposl, bfcmt below
    0.66666667 0.66666666 0.01380000 0.00000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.98620000 0.00000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.48620000 0.00000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.51380000 0.00000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.30710000 0.00000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.69290000 0.00000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.19290000 0.00000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.80710000 0.00000000 0.00000000 0.00000000
    'Nb.in' : spfname
    4 : natoms; atposl, bfcmt below
    0.00000000 0.00000000 0.35780000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.64220000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.14220000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.85780000 0.00000000 0.00000000 0.00000000
    'O.in' : spfname
    18 : natoms; atposl, bfcmt below
    0.00000000 0.29700000 0.25000000 0.00000000 0.00000000 0.00000000
    0.00000000 0.70300000 0.75000000 0.00000000 0.00000000 0.00000000
    0.29700000 0.70300000 0.25000000 0.00000000 0.00000000 0.00000000
    0.70300000 0.29700000 0.75000000 0.00000000 0.00000000 0.00000000
    0.70300000 0.00000000 0.25000000 0.00000000 0.00000000 0.00000000
    0.29700000 0.00000000 0.75000000 0.00000000 0.00000000 0.00000000
    0.30900000 0.03300000 0.08380000 0.00000000 0.00000000 0.00000000
    0.69100000 0.96700000 0.91620000 0.00000000 0.00000000 0.00000000
    0.03300000 0.65800000 0.08380000 0.00000000 0.00000000 0.00000000
    0.96700000 0.34200000 0.91620000 0.00000000 0.00000000 0.00000000
    0.65800000 0.30900000 0.08380000 0.00000000 0.00000000 0.00000000
    0.34200000 0.69100000 0.91620000 0.00000000 0.00000000 0.00000000
    0.34200000 0.96700000 0.41620000 0.00000000 0.00000000 0.00000000
    0.65800000 0.03300000 0.58380000 0.00000000 0.00000000 0.00000000
    0.69100000 0.34200000 0.41620000 0.00000000 0.00000000 0.00000000
    0.30900000 0.65800000 0.58380000 0.00000000 0.00000000 0.00000000
    0.96700000 0.69100000 0.41620000 0.00000000 0.00000000 0.00000000
    0.03300000 0.30900000 0.58380000 0.00000000 0.00000000 0.00000000

    Best,
    Zhiwei

     
  • Zhiwei Li

    Zhiwei Li - 2021-11-06

    no one have any idea......please help!

     
  • mfechner

    mfechner - 2021-11-06

    May I ask how your AFM setting looks like ?

    best
    Michael

     
  • mfechner

    mfechner - 2021-11-06

    May I ask how your AFM setting looks like ?

    best
    Michael

     
  • Zhiwei Li

    Zhiwei Li - 2021-11-06

    Dear Michael,

    Thank you for your interest and my AFM setting :
    This calculation converges after 34 loops on my server

    tasks
    0

    xctype
    3

    mixtype
    3

    !broydpm
    !0.3 0.1

    ! DFT+U block
    dft+u
    1 4 : dftu, inpdftu
    1 2 3.0 : is, l, lambda

    spinpol
    .true.

    ! external field for breaking spin symmetry
    bfieldc
    0.0 0.0 0.0

    reducebf
    0.8

    ! conduction local-orbitals required
    lorbcnd
    .true.

    ngridk
    4 4 4

    nempty
    8

    msmooth
    2

    lmaxi
    2

    maxscl
    500

    sppath
    './'

    avec
    4.894400000 8.477349473 0.000000000
    9.788800000 0.000000000 0.000000000
    0.000000000 0.000000000 26.76420000

    atoms
    3 : nspecies
    'Co.in' : spfname
    8 : natoms; atposl, bfcmt below
    0.66666667 0.66666666 0.01800000 0.010000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.98200000 -0.010000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.48200000 -0.010000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.51800000 0.010000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.30680000 0.010000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.69320000 -0.010000000 0.00000000 0.00000000
    0.33333333 0.33333334 0.19320000 -0.010000000 0.00000000 0.00000000
    0.66666667 0.66666666 0.80680000 0.010000000 0.00000000 0.00000000
    'Nb.in' : spfname
    4 : natoms; atposl, bfcmt below
    0.00000000 0.00000000 0.35720000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.64280000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.14280000 0.00000000 0.00000000 0.00000000
    0.00000000 0.00000000 0.85720000 0.00000000 0.00000000 0.00000000
    'O.in' : spfname
    18 : natoms; atposl, bfcmt below
    0.00000000 0.29600000 0.25000000 0.00000000 0.00000000 0.00000000
    0.00000000 0.70400000 0.75000000 0.00000000 0.00000000 0.00000000
    0.29600000 0.70400000 0.25000000 0.00000000 0.00000000 0.00000000
    0.70400000 0.29600000 0.75000000 0.00000000 0.00000000 0.00000000
    0.70400000 0.00000000 0.25000000 0.00000000 0.00000000 0.00000000
    0.29600000 0.00000000 0.75000000 0.00000000 0.00000000 0.00000000
    0.30300000 0.04300000 0.08420000 0.00000000 0.00000000 0.00000000
    0.69700000 0.95700000 0.91580000 0.00000000 0.00000000 0.00000000
    0.04300000 0.65400000 0.08420000 0.00000000 0.00000000 0.00000000
    0.95700000 0.34600000 0.91580000 0.00000000 0.00000000 0.00000000
    0.65400000 0.30300000 0.08420000 0.00000000 0.00000000 0.00000000
    0.34600000 0.69700000 0.91580000 0.00000000 0.00000000 0.00000000
    0.34600000 0.95700000 0.41580000 0.00000000 0.00000000 0.00000000
    0.65400000 0.04300000 0.58420000 0.00000000 0.00000000 0.00000000
    0.69700000 0.34600000 0.41580000 0.00000000 0.00000000 0.00000000
    0.30300000 0.65400000 0.58420000 0.00000000 0.00000000 0.00000000
    0.95700000 0.69700000 0.41580000 0.00000000 0.00000000 0.00000000
    0.04300000 0.30300000 0.58420000 0.00000000 0.00000000 0.00000000

    best
    Zhiwei
    
     
  • Zhiwei Li

    Zhiwei Li - 2022-06-16

    Is there anyone can help me to work this problem out.
    After many try out, I really can not get this non-magnetic + DFT+U calculation converged.

    Please help!

    Best,
    Zhiwei

     
  • Mike B

    Mike B - 2022-11-11

    Dear Zhiwei,

    I also did also some calculations for a non-magnetic material, including dft+u. I didn't had problems with the convergence, but rather that the DOS had some unexpected features compared to the results of other code (e.g. VASP), when I used spin unpolarized settings.

    Thats why, I did some test calculations for some (non-magnetic!) fcc Noble-metals (Cu, Ag, Rh, Pd, Au) and tested the influence of some Hubbard-U value in both double-counting schemes (AMF/FLL). As it turns out I get this weird results in the case of spinpol=.false. in addition with dft+u in comparison to the "correct" results (compared to other codes) of the DOS in the case of spinpol=.true. and dft+u.
    I suspect that this could be due to a bug, although I didn't found any obvious bug in the subroutine.

    Maybe this issue leads in your case to convergence problems.
    I would suggest to use spin polarization in combiantion with dft+u and to converge to a non-magnetic solution (maybe also consider using a fixed spin moment calculation with zero magnetization).

    Best regards,
    Mike

     
    • Zhiwei Li

      Zhiwei Li - 2022-11-12

      Thanks Mike, I will take your advice and do the spin polarized calculation.

      cheers,
      Zhiwei

       

Log in to post a comment.