[CP2K-user] cutoff-convergence

[Travis]

Hi,

The tutorial works because it is a simple case with small exponents in the 
basis set. If you have an element (or especially cation) with large 
exponents in the basis set, you need to use a higher cutoff. The CUTOFF 
should be at least as large as the largest exponent in your basis set 
multiplied by the REL_CUTOFF. Then round up to a nice number to report in 
the paper, nearest 10, 20, 50, or 100 are common. Na, Mg, H, and O are 
common elements to base your cutoff on.

Na DZVP-GTH-q9 DZVP-GTH
  2
  2  0  1  6  3  2
       *17.6698382207*  -0.1362364777   0.0270957219   0.0000000000   
0.0791980974   0.0000000000
        6.9070658063  -0.0645835666   0.0098734119   0.0000000000   
0.2184139257   0.0000000000
        2.6098601451   0.3866153961  -0.0737170396   0.0000000000   
0.3764887364   0.0000000000
        0.9649548796   0.5971358883  -0.2186742776   0.0000000000   
0.3922304007   0.0000000000
        0.3400348369   0.1756697026  -0.0301066303   0.0000000000   
0.1808407220   0.0000000000
        0.0432486046  -0.0017104621   1.0388865857   1.0000000000   
0.0057696324   1.0000000000
  3  2  2  1  1
        0.0973000000   1.0000000000

In my experience, REL_CUTOFF=50 Ry is a good choice for static cell 
calculations and REL_CUTOFF=80 Ry is a good choice for dynamic cell 
calculations. Both figures are based on experience with condensed phase 
simulation. This leads to CUTOFFs of ~600 Ry and ~1000 Ry for liquid water 
for NVT and NPT simulations, respectively. The devs had a recent paper on 
accurate simulations of liquid water which draws similar conclusions when 
not using smoothing.

-T


On Thursday, October 10, 2019 at 8:13:29 AM UTC-3, Akshay Malik wrote:
>
> Dear all,
> I am trying to do cutoff-convergence for my system but it is only 
> converging up to two digits after the decimal point.
> Please suggest some method by which I can get converged energy up to 6 
> decimals.  
>
> # Grid cutoff vs total energy
> # Date: Thu Oct 10 16:37:32 IST 2019
> # PWD: /home/space/cp2k/cutoff-convergence
> # REL_CUTOFF = 60
> # Cutoff (Ry) | Total Energy (Ha) | NG on grid 1 | NG on grid 2 | NG on 
> grid 3 | NG on grid 4
>      50.00  -1337.8915343935  668903   50882     240       0
>     100.00  -1351.0266889654  1216132  286203   11290       0
>     150.00  -1351.6349634609  1071797  364690   75938    1200
>     200.00  -1351.6825629692  975942  418150  115933    3600
>     250.00  -1351.6992612267  875755  439820  192650    5400
>     300.00  -1351.6990099247  791971  424161  286203   11290
>     350.00  -1351.6961310939  735698  431201  323555   23171
>     400.00  -1351.6944391280  687700  428173  341928   55824
>     450.00  -1351.6933108152  644645  427152  364690   77138
>     500.00  -1351.6931016574  603223  441589  368145  100668
>
> I am using this cp2k file section for calculating energy:
> &FORCE_EVAL
>   METHOD Quickstep
>   &DFT
>     BASIS_SET_FILE_NAME  GTH_BASIS_SETS
>     POTENTIAL_FILE_NAME  GTH_POTENTIALS
>     &MGRID
>       NGRIDS 4
>       CUTOFF LT_cutoff
>       REL_CUTOFF LT_rel_cutoff
>     &END MGRID
>     &QS
>       EPS_DEFAULT 1.0E-10
>     &END QS
>     &SCF
>       SCF_GUESS ATOMIC
>       EPS_SCF 1.0E-6
>       MAX_SCF 1
>       ADDED_MOS 10
>       CHOLESKY INVERSE
>       &SMEAR ON
>         METHOD FERMI_DIRAC
>         ELECTRONIC_TEMPERATURE [K] 333
>       &END SMEAR
>       &DIAGONALIZATION
>         ALGORITHM STANDARD
>       &END DIAGONALIZATION
>       &MIXING
>         METHOD BROYDEN_MIXING
>         ALPHA 0.4
>         BETA 0.5
>         NBROYDEN 8
>       &END MIXING
>     &END SCF
>      &XC
>       &XC_FUNCTIONAL BLYP
>       &END
>     &END
>   &END DFT
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20191010/4ce2947c/attachment.htm>