[CP2K-user] SCF does not converge for 1st step
Jiapeng Liu
sunta... at gmail.com
Sat Oct 12 00:46:08 UTC 2019
Hi Travis,
I am wondering if there are some parameters can help accelerate the scf
convergence? As I use the above setting with OT, after several steps the
scf becomes difficult to converge even if I set the SCF_EPS to 1E-5. Is it
fine to continue or are there any options. Thanks.
Regards,
Jiapeng
On Saturday, October 12, 2019 at 12:49:48 AM UTC+8, Travis wrote:
>
> Hi,
>
> For an insulator, you do not need kpoints (supposing you use a supercell
> with sides of 1 nm or greater), smearing, or mixing. Though these things
> shouldn't prevent convergence. You only listed Li in &KINDS, so I thought
> you were still doing calculations on the metal. You can try the OT method
> as well by deleting the mixing and kpoint sections in your input and
> swapping from standard to OT in the diagonalization section.
>
> &SCF
> &DIAGONALIZATION
> ALGORITHM OT
> &END DIAGONALIZATION
> &END SCF
>
> -T
>
> On Friday, October 11, 2019 at 1:32:03 PM UTC-3, Jiapeng Liu wrote:
>>
>> Hi Travis,
>>
>> Thanks for your suggestion, but it does not work. The SCF does not
>> converge even after 150 steps. May I ask why should we switch SMEAR on? I
>> thought it is necessary for metal, but here is an insulator, it is Li
>> conducting but not electron conductive. Thanks for your help.
>>
>> Regards,
>> Jiapeng
>>
>> On Saturday, October 12, 2019 at 12:08:38 AM UTC+8, Travis wrote:
>>>
>>> Hi,
>>>
>>> &SCF
>>> ...
>>> &SMEAR T
>>> METHOD FERMI_DIRAC
>>> ELECTRONIC_TEMPERATURE 300
>>> &END SMEAR
>>> ...
>>> &END SCF
>>>
>>> -T
>>>
>>> On Friday, October 11, 2019 at 12:58:02 PM UTC-3, Jiapeng Liu wrote:
>>>>
>>>> For your reference, the basis set I used is DZVP-MOLOPT-SR-GTH and
>>>> potential is GTH-PBE. Thanks for any help.
>>>>
>>>> On Friday, October 11, 2019 at 11:54:39 PM UTC+8, Jiapeng Liu wrote:
>>>>>
>>>>> Dear CP2K users,
>>>>>
>>>>> I am just learning CP2k now and I wanna use CP2K to optimize a Li
>>>>> solid state electrolyte. The problem is that the SCF run does not converge
>>>>> for the first step even after 180 scf steps. I am wondering if this is ok
>>>>> for the continue calculations or there is something wrong with my parameter
>>>>> setting. Can you kindly help me have a look. The input file is like
>>>>> following
>>>>>
>>>>> &GLOBAL
>>>>> PROJECT_NAME relax
>>>>> RUN_TYPE GEO_OPT
>>>>> PRINT_LEVEL LOW
>>>>> &END GLOBAL
>>>>>
>>>>> &MOTION
>>>>> &GEO_OPT
>>>>> MAX_ITER 200
>>>>> OPTIMIZER CG
>>>>> MAX_DR 1.0E-3
>>>>> MAX_FORCE 4.5E-4
>>>>> RMS_DR 1.0E-3
>>>>> RMS_FORCE 3.0E-4
>>>>> &END
>>>>> &END MOTION
>>>>>
>>>>> &FORCE_EVAL
>>>>> METHOD QS
>>>>> STRESS_TENSOR ANALYTICAL
>>>>> &DFT
>>>>> CHARGE 0
>>>>> MULTIPLICITY 1
>>>>> BASIS_SET_FILE_NAME BASIS_MOLOPT
>>>>> POTENTIAL_FILE_NAME GTH_POTENTIALS
>>>>>
>>>>> &MGRID
>>>>> NGRIDS 5
>>>>> CUTOFF 600
>>>>> REL_CUTOFF 80
>>>>> &END MGRID
>>>>>
>>>>> &QS
>>>>> METHOD GPW ! to optimize the geometry the GPW method will be used
>>>>> EPS_DEFAULT 1.0E-12
>>>>> EXTRAPOLATION USE_GUESS
>>>>> &END QS
>>>>>
>>>>> &SCF
>>>>> MAX_SCF 200
>>>>> EPS_SCF 1.0E-6
>>>>> SCF_GUESS RESTART
>>>>> ADDED_MOS 200
>>>>> &MIXING T
>>>>> METHOD BROYDEN_MIXING
>>>>> ALPHA 0.4
>>>>> NMIXING 5
>>>>> NBUFFER 8
>>>>> &END MIXING
>>>>> &END SCF
>>>>>
>>>>> &XC
>>>>> FUNCTIONAL_ROUTINE NEW
>>>>> DENSITY_CUTOFF 1.0E-12
>>>>> GRADIENT_CUTOFF 1.0E-12
>>>>> TAU_CUTOFF 1.0E-12
>>>>> &XC_FUNCTIONAL
>>>>> &PBE ! PBE exchange-correlation functional
>>>>> PARAMETRIZATION ORIG
>>>>> &END PBE
>>>>> &END XC_FUNCTIONAL
>>>>> &END XC
>>>>> &POISSON
>>>>> POISSON_SOLVER PERIODIC
>>>>> PERIODIC XYZ
>>>>> &END POISSON
>>>>> &KPOINTS
>>>>> SCHEME GAMMA 2 2 2
>>>>> FULL_GRID .TRUE.
>>>>> &END KPOINTS
>>>>> &END DFT
>>>>>
>>>>> &SUBSYS
>>>>> &CELL
>>>>> ABC .....
>>>>> ALPHA_BETA_GAMMA ..........
>>>>> PERIODIC XYZ
>>>>> MULTIPLE_UNIT_CELL .....
>>>>> &END CELL
>>>>> &COORD
>>>>> ...........
>>>>> &END COORD
>>>>>
>>>>> &KIND Li
>>>>> ............
>>>>> &END KIND
>>>>> &END SUBSYS
>>>>> &PRINT
>>>>> &FORCES ON
>>>>> &END FORCES
>>>>> &END PRINT
>>>>> &END FORCE_EVAL
>>>>>
>>>>> and the first scf run is shown like this
>>>>>
>>>>> SCF WAVEFUNCTION OPTIMIZATION
>>>>>
>>>>> Step Update method Time Convergence Total
>>>>> energy Change
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> 1 NoMix/Diag. 0.40E+00 2.2 33.06169264 -1105.3440392037
>>>>> -1.11E+03
>>>>> 2 Broy./Diag. 0.40E+00 2.5 18.58674968 -1325.2412363377
>>>>> -2.20E+02
>>>>> 3 Broy./Diag. 0.40E+00 2.5 13.64590095 -1613.5521238999
>>>>> -2.88E+02
>>>>> 4 Broy./Diag. 0.40E+00 2.5 6.33206879 -1767.5043485819
>>>>> -1.54E+02
>>>>> 5 Broy./Diag. 0.40E+00 2.5 3.10815819 -1788.5141959477
>>>>> -2.10E+01
>>>>> 6 Broy./Diag. 0.40E+00 2.5 3.46033998 -1840.7455766805
>>>>> -5.22E+01
>>>>> 7 Broy./Diag. 0.40E+00 2.5 2.74195755
>>>>> -1831.3589058361 9.39E+00
>>>>> ......................
>>>>> 196 Broy./Diag. 0.40E+00 2.7 1.63271069
>>>>> -1923.1236698860 6.48E+00
>>>>> 197 Broy./Diag. 0.40E+00 2.7 1.72647036 -1924.2660229966
>>>>> -1.14E+00
>>>>> 198 Broy./Diag. 0.40E+00 2.6 1.75474955
>>>>> -1920.5269108168 3.74E+00
>>>>> 199 Broy./Diag. 0.40E+00 2.6 1.57890425 -1926.1904852274
>>>>> -5.66E+00
>>>>> 200 Broy./Diag. 0.40E+00 2.6 1.67473633 -1926.3874701171
>>>>> -1.97E-01
>>>>>
>>>>> Thanks very much.
>>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20191011/947ef097/attachment.htm>
More information about the CP2K-user
mailing list